Package 'earth'

Title: Multivariate Adaptive Regression Splines
Description: Build regression models using the techniques in Friedman's papers "Fast MARS" and "Multivariate Adaptive Regression Splines" <doi:10.1214/aos/1176347963>. (The term "MARS" is trademarked and thus not used in the name of the package.)
Authors: Stephen Milborrow [aut, cre], Trevor Hastie [aut], Rob Tibshirani [aut], Alan Miller [ctb], Thomas Lumley [ctb]
Maintainer: Stephen Milborrow <[email protected]>
License: GPL-3
Version: 5.3.4
Built: 2024-11-05 03:41:15 UTC
Source: https://github.com/cran/earth

Help Index


Please ignore

Description

Contrasts function for factors in the earth response. For internal use by earth.

Usage

contr.earth.response(x, base, contrasts)

Arguments

x

a factor

base

unused

contrasts

unused

Value

Returns a diagonal matrix. An example for a 3 level factor with levels A, B, and C:

      A B C
    A 1 0 0
    B 0 1 0
    C 0 0 1

Note

Earth uses this function internally. You shouldn't need it. It is made publicly available only because it seems that is necessary for model.matrix.

See Also

contrasts


Multivariate Adaptive Regression Splines

Description

Build a regression model using the techniques in Friedman's papers "Multivariate Adaptive Regression Splines" and "Fast MARS".

See the package vignette “Notes on the earth package”.

Usage

## S3 method for class 'formula'
earth(formula = stop("no 'formula' argument"), data = NULL,
   weights = NULL, wp = NULL, subset = NULL,
   na.action = na.fail,
   pmethod = c("backward", "none", "exhaustive", "forward", "seqrep", "cv"),
   keepxy = FALSE, trace = 0, glm = NULL, degree = 1, nprune = NULL,
   nfold=0, ncross=1, stratify=TRUE,
   varmod.method = "none", varmod.exponent = 1,
   varmod.conv = 1, varmod.clamp = .1, varmod.minspan = -3,
   Scale.y = NULL, ...)

## Default S3 method:
earth(x = stop("no 'x' argument"), y = stop("no 'y' argument"),
    weights = NULL, wp = NULL, subset = NULL,
    na.action = na.fail,
    pmethod = c("backward", "none", "exhaustive", "forward", "seqrep", "cv"),
    keepxy = FALSE, trace = 0, glm = NULL, degree = 1, nprune = NULL,
    nfold=0, ncross=1, stratify=TRUE,
    varmod.method = "none", varmod.exponent = 1,
    varmod.conv = 1, varmod.clamp = .1, varmod.minspan = -3,
    Scale.y = NULL, ...)

## S3 method for class 'fit'
earth(x = stop("no 'x' argument"), y = stop("no 'y' argument"),
    weights = NULL, wp = NULL, subset = NULL,
    na.action = na.fail, offset = NULL,
    pmethod = c("backward", "none", "exhaustive", "forward", "seqrep", "cv"),
    keepxy = FALSE, trace = 0, glm = NULL, degree = 1,
    penalty = if(degree > 1) 3 else 2,
    nk = min(200, max(20, 2 * ncol(x))) + 1,
    thresh = 0.001, minspan = 0, endspan = 0,
    newvar.penalty = 0, fast.k = 20, fast.beta = 1,
    linpreds = FALSE, allowed = NULL,
    nprune = NULL, Object = NULL,
    Scale.y = NULL, Adjust.endspan = 2, Auto.linpreds = TRUE,
    Force.weights = FALSE, Use.beta.cache = TRUE, Force.xtx.prune = FALSE,
    Get.leverages = NROW(x) < 1e5, Exhaustive.tol = 1e-10, ...)

Arguments

To start off, look at the arguments formula, data, x, y, nk, degree, and trace.
If the response is binary or a factor, consider using the glm argument.
For cross validation, use the nfold argument.
For prediction intervals, use the varmod.method argument.

Most users will find that the above arguments are all they need, plus in some cases keepxy and nprune. Unless you are a knowledgeable user, it's best not subvert the standard algorithm by toying with tuning parameters such as thresh, penalty, and endspan.

formula

Model formula.

data

Data frame for formula.

x

Matrix or dataframe containing the independent variables.

y

Vector containing the response variable, or, in the case of multiple responses, a matrix or dataframe whose columns are the values for each response.

subset

Index vector specifying which cases to use, i.e., which rows in x to use. Default is NULL, meaning all.

weights

Case weights. Default is NULL, meaning no case weights. If specified, weights must have length equal to nrow(x) before applying subset. Zero weights are converted to a very small nonzero value. In the current implementation, building models with weights can be slow.

wp

Response weights. Default is NULL, meaning no response weights. If specified, wp must have an element for each column of y (after factors in y, if any, have been expanded). Zero weights are converted to a very small nonzero value.

na.action

NA action. Default is na.fail, and only na.fail is supported.

offset

Offset term passed from the formula in earth.formula.

keepxy

Default is FALSE. Set to TRUE to retain the following in the returned value: x and y (or data), subset, and weights. The function update.earth and friends will use these if present instead of searching for them in the environment at the time update.earth is invoked.
When the nfold argument is used with keepxy=TRUE, earth keeps more data and calls predict.earth multiple times to generate cv.oof.rsq.tab and cv.infold.rsq.tab (see the cv. arguments in earth.object). It therefore makes cross-validation significantly slower.

trace

Trace earth's execution. Values:
0 (default) no tracing
.3 variance model (the varmod.method arg)
.5 cross validation (the nfold arg)
1 overview
2 forward pass
3 pruning
4 model mats summary, pruning details
5 full model mats, internal details of operation

glm

NULL (default) or a list of arguments to pass on to glm. See the documentation of glm for a description of these arguments See “Generalized linear models” in the vignette. Example:
earth(survived~., data=etitanic, degree=2, glm=list(family=binomial))

The following arguments are for the forward pass.

degree

Maximum degree of interaction (Friedman's mimi). Default is 1, meaning build an additive model (i.e., no interaction terms).

penalty

Generalized Cross Validation (GCV) penalty per knot. Default is if(degree>1) 3 else 2. Simulation studies suggest values in the range of about 2 to 4. The FAQ section in the vignette has some information on GCVs.
Special values (for use by knowledgeable users): The value 0 penalizes only terms, not knots. The value -1 means no penalty, so GCV = RSS/n.

nk

Maximum number of model terms before pruning, i.e., the maximum number of terms created by the forward pass. Includes the intercept.
The actual number of terms created by the forward pass will often be less than nk because of other stopping conditions. See “Termination conditions for the forward pass” in the vignette.
The default is semi-automatically calculated from the number of predictors but may need adjusting.

thresh

Forward stepping threshold. Default is 0.001. This is one of the arguments used to decide when forward stepping should terminate: the forward pass terminates if adding a term changes RSq by less than thresh. See “Termination conditions for the forward pass” in the vignette.

minspan

Minimum number of observations between knots. (This increases resistance to runs of correlated noise in the input data.)
The default minspan=0 is treated specially and means calculate the minspan internally, as per Friedman's MARS paper section 3.8 with alphaalpha = 0.05. Set trace>=2 to see the calculated value.
Use minspan=1 and endspan=1 to consider all x values.
Negative values of minspan specify the maximum number of knots per predictor. These will be equally spaced. For example, minspan=-3 allows three evenly spaced knots for each predictor. As always, knots that fall in the end zones specified by endspan will be ignored.

endspan

Minimum number of observations before the first and after the final knot.
The default endspan=0 is treated specially and means calculate the endspan internally, as per the MARS paper equation 45 with alphaalpha = 0.05. Set trace>=2 to see the calculated value.
Be wary of reducing endspan, especially if you plan to make predictions beyond or near the limits of the training data. Overfitting near the edges of training data is much more likely with a small endspan. The model's RSq and GRSq won't indicate when this overfitting is occurring. (A plotmo plot can help: look for sharp hinges at the edges of the data). See also the Adjust.endspan argument.

newvar.penalty

Penalty for adding a new variable in the forward pass (Friedman's gammagamma, equation 74 in the MARS paper). Default is 0, meaning no penalty for adding a new variable. Useful non-zero values typically range from about 0.01 to 0.2 and sometimes higher — you will need to experiment.
A word of explanation. With the default newvar.penalty=0, if two variables have nearly the same effect (e.g. they are collinear), at any step in the forward pass earth will arbitrarily select one or the other (depending on noise in the sample). Both variables can appear in the final model, complicating model interpretation. On the other hand with a non-zero newvar.penalty, the forward pass will be reluctant to add a new variable — it will rather try to use a variable already in the model, if that does not affect RSq too much. The resulting final model may be easier to interpret, if you are lucky. There will often be a small performance hit (a worse GCV).

fast.k

Maximum number of parent terms considered at each step of the forward pass. (This speeds up the forward pass. See the Fast MARS paper section 3.0.)
Default is 20. A value of 0 is treated specially (as being equivalent to infinity), meaning no Fast MARS. Typical values, apart from 0, are 20, 10, or 5.
In general, with a lower fast.k (say 5), earth is faster; with a higher fast.k, or with fast.k disabled (set to 0), earth builds a better model. However, because of random variation this general rule often doesn't apply.

fast.beta

Fast MARS ageing coefficient, as described in the Fast MARS paper section 3.1. Default is 1. A value of 0 sometimes gives better results.

linpreds

Index vector specifying which predictors should enter linearly, as in lm. The default is FALSE, meaning predictors enter in the standard MARS fashion, i.e., in hinge functions.

The linpreds argument does not specify that a predictor must enter the model; only that if it enters, it enters linearly. See “The linpreds argument” in the vignette.
See also the Auto.linpreds argument below (which describes how earth will automatically treat a predictor as linear under certain conditions).

Details: A predictor's index in linpreds is the column number in the input matrix x (after factors have been expanded).
linpreds=TRUE makes all predictors enter linearly (the TRUE gets recycled).
linpreds may be a character vector e.g. linpreds=c("wind", "vis"). Note: grep is used for matching. Thus "wind" will match all variables that have "wind" in their names. Use "^wind$" to match only the variable named "wind".

allowed

Function specifying which predictors can interact and how. Default is NULL, meaning all standard MARS terms are allowed.
During the forward pass, earth calls the allowed function before considering a term for inclusion; the term can go into the model only if the allowed function returns TRUE. See “The allowed argument” in the vignette.

The following arguments are for the pruning pass.

pmethod

Pruning method. One of: backward none exhaustive forward seqrep cv.
Default is "backward".
Specify pmethod="cv" to use cross-validation to select the number of terms. This selects the number of terms that gives the maximum mean out-of-fold RSq on the fold models. Requires the nfold argument.
Use "none" to retain all the terms created by the forward pass.
If y has multiple columns, then only "backward" or "none" is allowed.
Pruning can take a while if "exhaustive" is chosen and the model is big (more than about 30 terms). The current version of the leaps package used during pruning does not allow user interrupts (i.e., you have to kill your R session to interrupt; in Windows use the Task Manager or from the command line use taskkill).

nprune

Maximum number of terms (including intercept) in the pruned model. Default is NULL, meaning all terms created by the forward pass (but typically not all terms will remain after pruning). Use this to enforce an upper bound on the model size (that is less than nk), or to reduce exhaustive search time with pmethod="exhaustive".

The following arguments are for cross validation.

nfold

Number of cross-validation folds. Default is 0, no cross validation. If greater than 1, earth first builds a standard model as usual with all the data. It then builds nfold cross-validated models, measuring R-Squared on the out-of-fold (left out) data each time. The final cross validation R-Squared (CVRSq) is the mean of these out-of-fold R-Squareds.
The above process of building nfold models is repeated ncross times (by default, once). Use trace=.5 to trace cross-validation.
Further statistics are calculated if keepxy=TRUE or if a binomial or poisson model (specified with the glm argument). See “Cross validation” in the vignette.

ncross

Only applies if nfold>1. Number of cross-validations. Each cross-validation has nfold folds. Default 1.

stratify

Only applies if nfold>1. Default is TRUE. Stratify the cross-validation samples so that an approximately equal number of cases with a non-zero response occur in each cross validation subset. So if the response y is logical, the TRUEs will be spread evenly across folds. And if the response is a multilevel factor, there will be an approximately equal number of each factor level in each fold (because a multilevel factor response gets expanded to columns of zeros and ones, see “Factors” in the vignette). We say “approximately equal” because the number of occurrences of a factor level may not be exactly divisible by the number of folds.

The following arguments are for variance models.

varmod.method

Construct a variance model. For details, see varmod and the vignette “Variance models in earth”. Use trace=.3 to trace construction of the variance model.
This argument requires nfold and ncross. (We suggest at least ncross=30 here to properly calculate the variance of the errors — although you can use a smaller value, say 3, for debugging.)
The varmod.method argument should be one of
"none" Default. Don't build a variance model.
"const" Assume homoscedastic errors.
"lm" Use lm to estimate standard deviation as a function of the predicted response.
"rlm" Use rlm.
"earth" Use earth.
"gam" Use gam. This will use either gam or the mgcv package, whichever is loaded.
"power" Estimate standard deviation as intercept + coef * predicted.response^exponent, where intercept, coef, and exponent will be estimated by nls. This is equivalent to varmod.method="lm" except that exponent is automatically estimated instead of being held at the value set by the varmod.exponent argument.
"power0" Same as "power" but no intercept (offset) term.
"x.lm", "x.rlm", "x.earth", "x.gam" Like the similarly named options above, but estimate standard deviation by regressing on the predictors x (instead of the predicted response). A current implementation restriction is that "x.gam" allows only models with one predictor (x must have only one column).

varmod.exponent

Power transform applied to the rhs before regressing the absolute residuals with the specified varmod.method. Default is 1.
For example, with varmod.method="lm", if you expect the standard deviance to increase linearly with the mean response, use varmod.exponent=1. If you expect the standard deviance to increase with the square root of the mean response, use varmod.exponent=.5 (where negative response values will be treated as 0, and you will get an error message if more than 20% of them are negative).

varmod.conv

Convergence criterion for the Iteratively Reweighted Least Squares used when creating the variance model.
Iterations stop when the mean value of the coefficients of the residual model change by less than varmod.conv percent. Default is 1 percent.
Negative values force the specified number of iterations, e.g. varmod.conv=-2 means iterate twice.
Positive values are ignored for varmod="const" and also currently ignored for varmod="earth" (these are iterated just once, the same as using varmod.conv=-1).

varmod.clamp

The estimated standard deviation of the main model errors is forced to be at least a small positive value, which we call min.sd. This prevents negative or absurdly small estimated standard deviations. Clamping takes place in predict.varmod, which is called by predict.earth when estimating prediction intervals. The value of min.sd is determined when building the variance model as min.sd = varmod.clamp * mean(sd(training.residuals)). The default varmod.clamp is 0.1.

varmod.minspan

Only applies when varmod.method="earth" or "x.earth". This is the minspan used in the internal call to earth when creating the variance model (not the main earth model).
Default is -3, i.e., three evenly spaced knots per predictor. Residuals tend to be very noisy, and allowing only this small number of knots helps prevent overfitting.

The following arguments are for internal or advanced use.

Object

Earth object to be updated, for use by update.earth.

Scale.y

Scale the response internally in the forward pass. Scaling here means subtract the mean and divide by the standard deviation.
For single-response models, the default is Scale.y = TRUE. Scaling is invisible to the user, up to numerical differences, but does provide better numeric stability.
For multiple-response models, the default is FALSE. If Scale.y is set TRUE, each column of the response is independently scaled. This can prevent one response from “overwhelming” the others, and earth typically generates a different set of hinge functions.

Adjust.endspan

In interaction terms, endspan gets multiplied by this value. This reduces the possibility of an overfitted interaction term supported by just a few cases on the boundary of the predictor space (as sometimes seen in our simulation studies).
The default is 2. Use Adjust.endspan=1 for compatibility with old versions of earth.

Auto.linpreds

Default is TRUE, which works as follows (see example):
At any step in the forward pass, if earth discovers that the best knot for the best predictor is at the predictor minimum (in the training data), then earth adds the predictor to the model as a linear “basis function” (with no hinge). Compare the following basis functions (printed in bold) for an example such predictor x:
Auto.linpreds=TRUE (default): x
Auto.linpreds=FALSE: max(x-99, 0) where 99 is the minimum x in the training data.
Using Auto.linpreds=FALSE always forces a knot, even when the knot is at the minimum value of the variable. This ensures that the basis functions are always expressed as hinge functions (and will always be non-negative).
Note that Auto.linpreds affects only how the model behaves outside the training data. Thus predict.earth will make the same predictions from the training data, regardless of whether the earth model was built with Auto.linpreds set TRUE or FALSE (up to possible differences in the size of the model caused by different GCVs because of the different forms of the terms).

Force.weights

Default is FALSE. For testing the weights argument. Force use of the code for handling weights in the earth code, even if weights=NULL or all the weights are the same. This will not necessarily generate an identical model, primarily because the non-weighted code requires some tests for numerical stability that can sometimes affect knot selection.

Use.beta.cache

Default is TRUE. Using the “beta cache” takes a little more memory but is faster (by 20% and often much more for large models). The beta cache uses nk * nk * ncol(x) * sizeof(double) bytes. (The beta cache is an innovation in this implementation of MARS and does not appear in Friedman's papers. It is not related to the fast.beta argument. Certain regression coefficients in the forward pass can be saved and re-used, thus saving recalculation time.)

Force.xtx.prune

Default is FALSE. This argument pertains to subset evaluation in the pruning pass. By default, if y has a single column then earth calls the leaps routines; if y has multiple columns then earth calls EvalSubsetsUsingXtx. The leaps routines are numerically more stable but do not support multiple responses (leaps is based on the QR decomposition and EvalSubsetsUsingXtx is based on the inverse of X'X). Setting Force.xtx.prune=TRUE forces use of EvalSubsetsUsingXtx, even if y has a single column.

Get.leverages

Default is TRUE unless the model has more than 100 thousand cases. The leverages are the diagonal hat values for the linear regression of y on bx. (The leverages are needed only for certain model checks, for example when plotres is called with versus=4).
Details: This argument was introduced to reduce peak memory usage. When n >> p, memory use peaks when earth is calculating the leverages.

Exhaustive.tol

Default 1e-10. Applies only when pmethod="exhaustive". If the reciprocal of the condition number of bx is less than Exhaustive.tol, earth forces pmethod="backward". See “XHAUST returned error code -999” in the vignette.

...

Dots are passed on to earth.fit.

Value

An S3 model of class "earth". See earth.object for a complete description.

Author(s)

Stephen Milborrow, derived from mda::mars by Trevor Hastie and Robert Tibshirani.

The approach used for GLMs was motivated by work done by Jane Elith and John Leathwick (a representative paper is given below).

The evimp function uses ideas from Max Kuhn's caret package https://CRAN.R-project.org/package=caret.

Parts of Thomas Lumley's leaps package have been incorporated into earth, so earth can directly access Alan Miller's Fortran functions without going through hidden functions in the leaps package.

References

The Wikipedia article is recommended for an elementary introduction. The primary references are the Friedman papers, but readers may find the MARS section in Hastie, Tibshirani, and Friedman a more accessible introduction. Faraway takes a hands-on approach, using the ozone data to compare mda::mars with other techniques. (If you use Faraway's examples with earth instead of mars, use $bx instead of $x, and check out the book's errata.) Friedman and Silverman is recommended background reading for the MARS paper. Earth's pruning pass uses code from the leaps package which is based on techniques in Miller.

Faraway (2005) Extending the Linear Model with R https://www.maths.bath.ac.uk/~jjf23

Friedman (1991) Multivariate Adaptive Regression Splines (with discussion) Annals of Statistics 19/1, 1–141 http://projecteuclid.org/euclid.aos/1176347963
doi:10.1214/aos/1176347963

Friedman (1993) Fast MARS Stanford University Department of Statistics, Technical Report 110 https://statistics.stanford.edu/research/fast-mars

Friedman and Silverman (1989) Flexible Parsimonious Smoothing and Additive Modeling Technometrics, Vol. 31, No. 1.

Hastie, Tibshirani, and Friedman (2009) The Elements of Statistical Learning (2nd ed.) https://hastie.su.domains/pub.htm

Leathwick, J.R., Rowe, D., Richardson, J., Elith, J., & Hastie, T. (2005) Using multivariate adaptive regression splines to predict the distributions of New Zealand's freshwater diadromous fish Freshwater Biology, 50, 2034-2052 https://hastie.su.domains/pub.htm

Miller, Alan (1990, 2nd ed. 2002) Subset Selection in Regression https://wp.csiro.au/alanmiller/index.html

Wikipedia article on MARS https://en.wikipedia.org/wiki/Multivariate_adaptive_regression_splines

See Also

Start with summary.earth, plot.earth, evimp, and plotmo.

Please see the main package vignette “Notes on the earth package”. The vignette can also be downloaded from http://www.milbo.org/doc/earth-notes.pdf.

The vignette “Variance models in earth” is also included with the package. It describes how to generate prediction intervals for earth models.

Examples

earth.mod <- earth(Volume ~ ., data = trees)
plotmo(earth.mod)
summary(earth.mod, digits = 2, style = "pmax")

An earth object

Description

The object returned by the earth function.

This is an S3 model of class "earth". It is a list with the components listed below.

Term refers to a term created during the forward pass (each line of the output from format.earth is a term). Term number 1 is always the intercept.

Value

rss

Residual sum-of-squares (RSS) of the model (summed over all responses, if y has multiple columns).

rsq

1-rss/tss. R-Squared of the model (calculated over all responses, and calculated using the weights argument if it was supplied). A measure of how well the model fits the training data. Note that tss is the total sum-of-squares, sum((y - mean(y))^2).

gcv

Generalized Cross Validation (GCV) of the model (summed over all responses). The GCV is calculated using the penalty argument. For details of the GCV calculation, see equation 30 in Friedman's MARS paper and earth:::get.gcv.

grsq

1-gcv/gcv.null. An estimate of the predictive power of the model (calculated over all responses, and calculated using the weights argument if it was supplied). gcv.null is the GCV of an intercept-only model. See “Can GRSq be negative?” in the vignette.

bx

Matrix of basis functions applied to x. Each column corresponds to a selected term. Each row corresponds to a row in in the input matrix x, after taking subset. See model.matrix.earth for an example of bx handling. Example bx:

     (Intercept)  h(Girth-12.9)  h(12.9-Girth)  h(Girth-12.9)*h(...
[1,]           1            0.0            4.6                    0
[2,]           1            0.0            4.3                    0
[3,]           1            0.0            4.1                    0
...
dirs

Matrix with one row per MARS term, and with with ij-th element equal to

0 if predictor j is not in term i
-1 if an expression of the form h(const - xj) is in term i
1 if an expression of the form h(xj - const) is in term i
2 if predictor j should enter term i linearly (either because specified by the linpreds argument or because earth discovered that a knot was unnecessary).

This matrix includes all terms generated by the forward pass, including those not in selected.terms. Note that here the terms may not all be in pairs, because although the forward pass add terms as hinged pairs (so both sides of the hinge are available as building blocks for further terms), it also deletes linearly dependent terms before handing control to the pruning pass. Example dirs:

                        Girth Height
(Intercept)                 0      0  # intercept
h(12.9-Girth)              -1      0  # 2nd term uses Girth
h(Girth-12.9)               1      0  # 3rd term uses Girth
h(Girth-12.9)*h(Height-76)  1      1  # 4th term uses Girth and Height
...
cuts

Matrix with ij-th element equal to the cut point (hinge value) for predictor j in term i. This matrix includes all terms generated by the forward pass, including those not in selected.terms. Note for programmers: the precedent is to use dirs for term names etc. and to only use cuts where cut information needed. Example cuts:

                           Girth Height
(Intercept)                    0      0  # intercept, no cuts
h(12.9-Girth)               12.9      0  # 2nd term has cut at 12.9
h(Girth-12.9)               12.9      0  # 3rd term has cut at 12.9
h(Girth-12.9)*h(Height-76)  12.9     76  # 4th term has two cuts
...
prune.terms

A matrix specifying which terms appear in which pruning pass subsets. The row index of prune.terms is the model size. (The model size is the number of terms in the model. The intercept is counted as a term.) Each row is a vector of term numbers for the best model of that size. An element is 0 if the term is not in the model, thus prune.terms is a lower triangular matrix, with dimensions nprune x nprune. The model selected by the pruning pass is at row number length(selected.terms).

Example prune.terms:

     [,1] [,2] [,3] [,4] [,5] [,6] [,7]
[1,]    1    0    0    0    0    0    0   # intercept-only model
[2,]    1    2    0    0    0    0    0   # best 2 term model uses terms 1,2
[3,]    1    2    4    0    0    0    0   # best 3 term model uses terms 1,2,4
[4,]    1    2    6    9    0    0    0   # and so on
...
selected.terms

Vector of term numbers in the selected model. Can be used as a row index vector into cuts and dirs. The first element selected.terms[1] is always 1, the intercept.

fitted.values

Fitted values. A matrix with dimensions nrow(y) x ncol(y) after factors in y have been expanded.

residuals

Residuals. A matrix with dimensions nrow(y) x ncol(y) after factors in y have been expanded.

coefficients

Regression coefficients. A matrix with dimensions length(selected.terms) x ncol(y) after factors in y have been expanded. Each column holds the least squares coefficients from regressing that column of y on bx. The first row holds the intercept coefficient(s).

rss.per.response

A vector of the RSS for each response. Length is the number of responses, i.e., ncol(y) after factors in y have been expanded. The rss component above is equal to sum(rss.per.response).

rsq.per.response

A vector of the R-Squared for each response (where R-Squared is calculated using the weights argument if it was supplied). Length is the number of responses.

gcv.per.response

A vector of the GCV for each response. Length is the number of responses. The gcv component above is equal to sum(gcv.per.response).

grsq.per.response

A vector of the GRSq for each response (calculated using the weights argument if it was supplied). Length is the number of responses.

rss.per.subset

A vector of the RSS for each model subset generated by the pruning pass. Length is nprune. For multiple responses, the RSS is summed over all responses for each subset. The rss above is
rss.per.subset[length(selected.terms)]. The RSS of an intercept only-model is rss.per.subset[1].

gcv.per.subset

A vector of the GCV for each model in prune.terms. Length is nprune. For multiple responses, the GCV is summed over all responses for each subset. The gcv above is gcv.per.subset[length(selected.terms)]. The GCV of an intercept-only model is gcv.per.subset[1].

leverages

Diagonal of the hat matrix (from the linear regression of the response on bx).

penalty, nk, thresh

Copies of the corresponding arguments to earth.

pmethod, nprune

Copies of the corresponding arguments to earth.

weights, wp

Copies of the corresponding arguments to earth.

termcond

Reason the forward pass terminated (an integer).

call

The call used to invoke earth.

terms

Model frame terms. This component exists only if the model was built using earth.formula.

modvars

A matrix specifying which input variables are used in each column of the model matrix. (This field is new in earth 5.2.0.)
Columns correspond to columns of the model matrix (same as cols of dirs, see above).
Rows correspond to variables in the formula.

For example, the formula:

    survived ~ age + pclass + sqrt(age) - sex

results in:

attr(terms,"factors"):

                age  pclass  sqrt(age)
    survived      0       0          0  # the response will be dropped
    age           1       0          0
    pclass        0       1          0
    sqrt(age)     0       0          1  # sqrt(age) will be merged with age
    sex           0       0          0  # sex is unused and will be dropped

modvars:

            age pclass2nd pclass3rd sqrt(age)
    age       1         0         0         1  # age and sqrt(age) use "age"
    pclass    0         1         1         0  # pclass2nd and pclass3rd use "pclass"

Note that for models built with earth.default (x,y models), “derived variables” are not combined in modvars as they are for formula models. The row names of modvars match the column names of x, after factor expansion. Columns in x named "age" and "sqrt(age)" will be treated as two separate variables.

namesx

Variable names in the input data. Deprecated (subsumed by modvars).

xlevels

This component exists only if the model was built using earth.formula.
Same as lm. A record of the levels of the factors used in fitting, needed under certain conditions by predict.earth.

levels

This component exists only if the model was built using earth.default.
Levels of y if y is a factor,
c(FALSE,TRUE) if y is logical,
Else NULL.

The following fields appear only if earth's argument keepxy is TRUE.

x, y, data, subset

Copies of the corresponding arguments to earth. Only exist if keepxy=TRUE.

The following fields appear only if earth's glm argument is used.

glm.list

List of GLM models. Each element is the value returned by earth's internal call to glm for each response.
Thus if there is a single response (or a single binomial pair, see “Binomial pairs” in the vignette) this will be a one element list and you access the GLM model with earth.mod$glm.list[[1]].

glm.coefficients

GLM regression coefficients. Analogous to the coefficients field described above but for the GLM model(s). A matrix with dimensions length(selected.terms) x ncol(y) after factors in y have been expanded. Each column holds the coefficients from the GLM regression of that column of y on bx. This duplicates, for convenience, information buried in glm.list.

glm.stats

GLM summary statistics such as devratio, AIC, and iters.

glm.bpairs

Is NULL unless there are paired binomial columns. Else a logical vector c(TRUE, FALSE). See “Binomial pairs” in the vignette. Retained for backwards compatibility with old versions of earth.

The following fields appear only if the nfold argument is greater than 1.

cv.list

List of earth models, one model for each fold (ncross * nfold models).
The fold models have two extra fields, icross (an integer from 1 to ncross) and ifold (an integer from 1 to nfold).
To save memory, lengthy fields in the fold models are removed unless you use keepxy=TRUE. The “lengthy fields” are $bx, $fitted.values, and $residuals.

cv.nterms

Vector of length ncross * nfold + 1. Number of MARS terms in the model generated at each cross-validation fold, with the final element being the mean of these.

cv.nvars

Vector of length ncross * nfold + 1. Number of predictors in the model generated at each cross-validation fold, with the final element being the mean of these.

cv.groups

Specifies which cases went into which folds. Matrix with two columns and number of rows equal to the the number of cases nrow(x) Elements of the first column specify the cross-validation number, 1:ncross. Elements of the second column specify the fold number, 1:nfold.

cv.rsq.tab

Matrix with ncross * nfold + 1 rows and nresponse+1 columns, where nresponse is the number of responses, i.e., ncol(y) after factors in y have been expanded. The first nresponse elements of a row are the cv.rsq's on the out-of-fold data for each response of the model generated at that row's fold. (A cv.rsq is calculated from predictions on the out-of-fold data using the best model built from the in-fold data; where “best” means the model was selected using the in-fold GCV. The R-Squareds are calculated using the weights argument if it was supplied. The final column holds the row mean (a weighted mean if wp if specified)). The final row holds the column means. The values in this final row is the mean cv.rsq printed by summary.earth.

Example for a single response model (where the mean column is redundant but included for uniformity with multiple response models):

           y  mean
fold1  0.909 0.909
fold2  0.869 0.869
fold3  0.952 0.952
fold4  0.157 0.157
fold5  0.961 0.961
mean   0.769 0.769

Example for a multiple response model:

         y1   y2    y3   mean
fold1 0.915 0.951 0.944 0.937
fold2 0.962 0.970 0.970 0.968
fold3 0.914 0.940 0.942 0.932
fold4 0.907 0.929 0.925 0.920
fold5 0.947 0.987 0.979 0.971
mean  0.929 0.955 0.952 0.946

cv.class.rate.tab

Like cv.rsq.tab but is the classification rate at each fold i.e. the fraction of classes correctly predicted. Models with discrete response only. Calculated with thresh=.5 for binary responses. For responses with more than two levels, the final row is the overall classification rate. The other rows are the classification rates for each level (the level versus not-the-level), which are usually higher than the overall classification rate (predicting the level versus not-the-level is easier than correctly predicting one of many levels). The weights argument is ignored for all cross-validation stats except R-Squareds.

cv.maxerr.tab

Like cv.rsq.tab but is the MaxErr at each fold. This is the signed max absolute value at each fold. Results are aggregated for the final column and final row using the signed max absolute value. The signed max absolute value is defined as the maximum of the absolute difference between the predicted and observed response values, multiplied by -1 if the sign of that difference is negative.

cv.auc.tab

Like cv.rsq.tab but is the AUC at each fold. Binomial models only.

cv.cor.tab

Like cv.rsq.tab but is the cor at each fold. Poisson models only.

cv.deviance.tab

Like cv.rsq.tab but is the MeanDev at each fold. Binomial models only.

cv.calib.int.tab

Like cv.rsq.tab but is the CalibInt at each fold. Binomial models only.

cv.calib.slope.tab

Like cv.rsq.tab but is the CalibSlope at each fold. Binomial models only.

cv.oof.rsq.tab

Generated only if keepxy=TRUE or pmethod="cv".
A matrix with ncross * nfold + 1 rows and max.nterms columns, Each element holds an out-of-fold RSq (oof.rsq), calculated from predictions from the out-of-fold observations using the model built with the in-fold data. The final row is the mean over all folds. The R-Squareds are calculated using the weights argument if it was supplied.

cv.infold.rsq.tab

Generated only if keepxy=TRUE. Like cv.oof.rsq.tab but from predictions made on the in-fold observations.

cv.oof.fit.tab

Generated only if the varmod.method argument is used. Predicted values on the out-of-fold data. Dataframe with nrow(data) rows and ncross columns.

The following field appears only if the varmod.method is specified.

varmod

An object of class "varmod". See the varmod help page for a description. Only appears if the varmod.method argument is used.

See Also

earth


Titanic data with incomplete cases removed

Description

Titanic data with incomplete cases, passenger names, and other details removed.

Format

A data frame with 1046 observations on 6 variables.

pclass passenger class, unordered factor: 1st 2nd 3rd
survived integer: 0 or 1
sex unordered factor: male female
age age in years, min 0.167 max 80.0
sibsp number of siblings or spouses aboard, integer: 0...8
parch number of parents or children aboard, integer: 0...6

Source

This dataset is included in the earth package because it is a convenient vehicle for illustrating earth's GLM and factor handling.

The dataset was compiled by Frank Harrell and Robert Dawson: https://hbiostat.org/data/repo/titanic.html

For this version of the Titanic data, passenger details and incomplete cases were deleted and the name changed to etitanic to minimize confusion with other versions ("e" because it is part of the earth package).

Note that survived is an integer (it should arguably be a logical).

In this data the crew are conspicuous by their absence.

Contents of etitanic:

         pclass survived    sex    age sibsp parch
    1       1st        1 female 29.000     0     0
    2       1st        1   male  0.917     1     2
    3       1st        0 female  2.000     1     2
    4       1st        0   male 30.000     1     2
    5       1st        0 female 25.000     1     2
    ...
    1309    3rd        0   male 29.000     0     0
    

How etitanic was built:

    load("titanic3") # from Harrell's web site
    # discard name, ticket, fare, cabin, embarked, body, home.dest
    etitanic <- titanic3[,c(1,2,4,5,6,7)]
    etitanic <- etitanic[!is.na(etitanic$age),]
    save(etitanic, file="etitanic.rda")
    

References

Further details and analyses of the Titanic data may be found in:

F. Harrell (2001) Regression Modeling Strategies with Applications to Linear Models, Logistic Regression, and Survival Analysis https://hbiostat.org/rmsc

See Also

earth


Estimate variable importances in an earth object

Description

Estimate variable importances in an earth object

Usage

evimp(object, trim=TRUE, sqrt.=TRUE)

Arguments

object

An earth object.

trim

If TRUE (default), delete rows in the returned matrix for variables that don't appear in any subsets.

sqrt.

Default is TRUE, meaning take the sqrt of the GCV and RSS importances before normalizing to 0 to 100. Taking the square root gives a better indication of relative importances because the raw importances are calculated using a sum of squares. Use FALSE to not take the square root.

Value

This function returns a matrix showing the relative importances of the variables in the model. There is a row for each variable. The row name is the variable name, but with -unused appended if the variable does not appear in the final model.

The columns of the matrix are (not all of these are printed by print.evimp):

  • col: Column index of the variable in the x argument to earth.

  • used: 1 if the variable is used in the final model, else 0. Equivalently, 0 if the row name has an -unused suffix.

  • nsubsets: Variable importance using the "number of subsets" criterion. Is the number of subsets that include the variable (see "Three Criteria" in the chapter on evimp in the earth vignette “Notes on the earth package”).

  • gcv: Variable importance using the GCV criterion (see "Three Criteria").

  • gcv.match: 1, except is 0 where the rank using the gcv criterion differs from that using the nsubsets criterion. In other words, there is a 0 for values that increase as you go down the gcv column.

  • rss: Variable importance using the RSS criterion (see "Three Criteria").

  • rss.match: Like gcv.match but for the rss.

The rows are sorted on the nsubsets criterion. This means that values in the nsubsets column decrease as you go down the column (more accurately, they are non-increasing). The values in the gcv and rss columns are also non-increasing, except where the gcv or rss rank differs from the nsubsets ranking.

Note

There is a chapter on evimp in the earth package vignette “Notes on the earth package”.

Acknowledgment

Thanks to Max Kuhn for the original evimp code and for helpful discussions.

See Also

earth, plot.evimp

Examples

data(ozone1)
earth.mod <- earth(O3 ~ ., data=ozone1, degree=2)
ev <- evimp(earth.mod, trim=FALSE)
plot(ev)
print(ev)

Expand binomial-pair data from short to long form

Description

Expand binomial-pair data from “short” to “long” form.

The short form specifies the response with two columns giving the numbers of successes and failures. Example short form:

    survived died dose    sex
           3    0   10   male
           2    1   10 female
           1    2   20   male
           1    2   20 female

The long form specifies the response as single column of TRUEs and FALSEs. For example, the long form of the above data (spaces and comments added):

    survived dose    sex
        TRUE   10   male     # row 1 of short data: 0 died, 3 survived
        TRUE   10   male
        TRUE   10   male

       FALSE   10 female     # row 2 of short data: 1 died, 2 survived
        TRUE   10 female
        TRUE   10 female

       FALSE   20   male     # row 3 of short data: 2 died, 1 survived
       FALSE   20   male
        TRUE   20   male

       FALSE   20 female     # row 4 of short data: 2 died, 1 survived
       FALSE   20 female
        TRUE   20 female

In this example the total number of survived and died for each row in the short data is the same, but in general that need not be true.

Usage

## S3 method for class 'formula'
expand.bpairs(formula = stop("no 'formula' argument"), data = NULL, sort = FALSE, ...)

## Default S3 method:
expand.bpairs(data = stop("no 'data' argument"), y = NULL, sort = FALSE, ...)

Arguments

formula

Model formula such as survived + died ~ dose + temp.

data

Matrix or dataframe containing the data.

y

Model response. One of:
o Two column matrix or dataframe of binomial pairs e.g. cbind(survived, died=20-survived)
o Two-element numeric vector specifying the response columns in data e.g. c(1,2)
o Two-element character vector specifying the response column names in data e.g. c("survived", "died"). The full names must be used (partial matching isn't supported).

sort

Default FALSE. Use TRUE to sort the rows of the long data so it is returned in canonical form, independent of the row order of the short data. The long data is sorted on predictor values; predictors on the left take precedence in the sort order.

...

Unused, but provided for generic/method consistency.

Value

A dataframe of the data in the long form, with expanded binomial pairs. The first column of the data will be the response column (a column of TRUEs and FALSEs).

Additionally, the returned value has two attached attributes:

bpairs.index A vector of row indices into the returned data. Can be used to reconstruct the short data from the long data (although this package does not yet provide a function to do so).

ynames Column names of the original response (a two-element character vector).

Examples

survived <- c(3,2,1,1) # short data for demo (too short to build a real model)
died     <- c(0,1,2,2)
dose <- c(10,10,20,20)
sex  <- factor(c("male", "female", "male", "female"))

short.data <- data.frame(survived, died, dose, sex)

expand.bpairs(survived + died ~ ., short.data) # returns long form of the data

# expand.bpairs(data=short.data, y=cbind(survived, died)) # equivalent
# expand.bpairs(short.data, c(1,2))                       # equivalent
# expand.bpairs(short.data, c("survived", "died"))        # equivalent

# For example models, see the earth vignette
# section "Short versus long binomial data".

Format earth objects

Description

Return a string representing an earth expression (summary.earth calls this function internally to display the terms of the earth model).

Usage

## S3 method for class 'earth'
format(x = stop("no 'x' argument"),
       style = "h", decomp = "anova", digits = getOption("digits"),
       use.names = TRUE, colon.char = ":", ...)

Arguments

x

An earth object. This is the only required argument.

style

Formatting style. One of
"h" (default) more compact
"pmax" for those who prefer it
"max" is the same as "pmax" but prints max rather than pmax
"C" C style expression with zero based indexing
"bf" basis function format

decomp

One of
"anova" (default) order the terms using the "anova decomposition", i.e., in increasing order of interaction
"none" order the terms as created during the earth forward pass.

digits

Number of significant digits. The default is getOption(digits).

use.names

One of
TRUE (default), use variable names if available.
FALSE use names of the form x[,1].

colon.char

Change colons in the returned string to colon.char. Default is ":" (no change). Specifying colon.char="*" can be useful in some contexts to change names of the form x1:x2 to x1*x2.

...

Unused, but provided for generic/method consistency.

Value

A character representation of the earth model.

If there are multiple responses, format.earth will return multiple strings.

If there are embedded GLM model(s), the strings for the GLM model(s) come after the strings for the standard earth model(s).

Note

The FAQ section in the package vignette gives precise details of the "anova" ordering.

Using format.earth, perhaps after hand editing the returned string, you can create an alternative to predict.earth. For example:

as.func <- function(object, digits = 8, use.names = FALSE, ...)
  eval(parse(text=paste(
    "function(x){\n",
    "if(is.vector(x))\n",
    "  x <- matrix(x, nrow = 1, ncol = length(x))\n",
    "with(as.data.frame(x),\n",
    format(object, digits = digits, use.names = use.names, style = "pmax", ...),
    ")\n",
    "}\n", sep = "")))

earth.mod <- earth(Volume ~ ., data = trees)
my.func <- as.func(earth.mod, use.names = FALSE)
my.func(c(10,80))             # returns 16.84
predict(earth.mod, c(10,80))  # returns 16.84

Note that with pmax the R expression generated by format.earth can handle multiple cases. Thus the expression is consistent with the way predict functions usually work in R — we can give predict multiple cases (i.e., multiple rows in the input matrix) and it will return a vector of predicted values.

 

The earth package also provides a function format.lm. It has arguments as follows
format.lm(x, digits=getOption("digits"), use.names=TRUE, colon.char=":")
(Strictly speaking, format.lm doesn't belong in the earth package.) Example:

lm.mod <- lm(Volume ~ Height*Girth, data = trees)
cat(format(lm.mod, colon.char="*"))

# yields:
#    69.4
#    -  1.30 * Height
#    -  5.86 * Girth
#    + 0.135 * Height*Girth

See Also

summary.earth, pmax,

Examples

earth.mod <- earth(Volume ~ ., data = trees)
cat(format(earth.mod))

# yields:
#    29.0
#    -  3.42 * h(14.2-Girth)
#    +  6.23 * h(Girth-14.2)
#    + 0.581 * h(Height-75)

cat(format(earth.mod, style="pmax"))

# yields:
#    29.0
#    -  3.42 * pmax(0,   14.2 -  Girth)
#    +  6.23 * pmax(0,  Girth -  14.2)
#    + 0.581 * pmax(0, Height -  75)

cat(format(earth.mod, style="C"))

# yields (note zero based indexing):
#    29.0
#    -  3.42 * max(0, 14.2 - x[0])
#    +  6.23 * max(0, x[0] - 14.2)
#    + 0.581 * max(0, x[1] - 75)

cat(format(earth.mod, style="bf"))

# yields:
#    29.0
#    -  3.42 * bf1
#    +  6.23 * bf2
#    + 0.581 * bf3
#
#     bf1  h(14.2-Girth)
#     bf2  h(Girth-14.2)
#     bf3  h(Height-75)

Convert a mars object from the mda package to an earth object

Description

Convert a mars object from the mda package to an earth object

Usage

mars.to.earth(object, trace=TRUE)

Arguments

object

A mars object, created using mars in the mda package.

trace

If TRUE (default) print a summary of the conversion.

Value

The value is the same format as that returned by earth but with skeletal versions of rss.per.subset, gcv.per.subset, and prune.terms.

You can fully initialize these components by calling update.earth after mars.to.earth, but if you do this selected.terms may change. However with pmethod="backward" a change is unlikely — selected.terms would change only if GCVs are so close that numerical errors have an effect.

Note

Differences between mars and earth objects

Perhaps the most notable difference between mars and earth objects is that mars returns the MARS basis matrix in a field called "x" whereas earth returns "bx" with only the selected terms. Also, earth returns "dirs" rather than "factors", and in earth this matrix can have entries of value 2 for linear predictors.

For details of other differences between mars and earth objects, see the comments in the source code of mars.to.earth.

Weights

The w argument is silently ignored by mars.

mars normalizes wp to (euclidean) length 1; earth normalizes wp to length equal to the number of responses, i.e., the number of columns in y. This change was made so an all ones wp (or in fact any all constant wp) is equivalent to using no wp.

If the original call to mars used the wp argument, mars.to.earth will run update.earth to force consistency. This could modify the model, so a warning is issued.

See Also

earth, mars

Examples

if(require(mda)) {
    mars.mod <- mars(trees[,-3], trees[,3])
    earth.mod <- mars.to.earth(mars.mod)
    # the standard earth functions can now be used
    # note the reconstructed call in the summary
    summary(earth.mod, digits = 2)
}

Get the earth basis matrix

Description

Get the basis matrix of an earth model.

Usage

## S3 method for class 'earth'
model.matrix(object = stop("no 'object' argument"),
    x = NULL, subset = NULL, which.terms = NULL,
    trace = 0,
    ...,
    Env = parent.frame(),
    Callers.name = "model.matrix.earth")

Arguments

object

An earth model. This is the only required argument.

x

Default is NULL, meaning use the original data used to build the earth model (after taking the original subset, if any).

Else x can be a data frame, a matrix, or a vector with length equal to a multiple of the number of columns of the original input matrix x. (There is some leniency here. For example, column names aren't necessary if x has the same number of predictors originally used to build the earth model.)

subset

Which rows to use in x. Default is NULL, meaning use all of x.

which.terms

Which terms to use. Default is NULL, meaning all terms in the earth model (i.e. the terms in object$selected.terms).

trace

Default 0. Set to non-zero to see which data model.matrix.earth is using.

...

Unused, but provided for generic/method consistency.

Env

For internal use.

Callers.name

For internal use (used by earth in trace messages).

Value

A basis matrix bx of the same form returned by earth. The format of bx is described in earth.object.

If x, subset, and which.terms are all NULL (the default), this function returns the model's bx. In this case, it is perhaps easier to simply use object$bx.

The matrix bx can be used as the input matrix to lm or glm, as shown below in the example. In fact, that is what earth does internally after the pruning pass — it calls lm.fit, and additionally glm if earth's glm argument is used.

See Also

earth

Examples

# Example 1

data(trees)
earth.mod <- earth(Volume ~ ., data = trees) # standard earth model
summary(earth.mod, decomp = "none")  # "none" to print terms in same order as lm.mod below
bx <- model.matrix(earth.mod)        # earth model's basis mat (equivalent to bx <- earth.mod$bx)
lm.mod <- lm(trees$Volume ~ bx[,-1]) # -1 to drop intercept
summary(lm.mod)                      # yields same coeffs as above summary
                                     # displayed t values are not meaningful

# Example 2

earth.mod <- earth(Volume~., data=trees) # standard earth model
summary(earth.mod, decomp = "none")  # "none" to print terms in same order as lm.mod below
bx <- model.matrix(earth.mod)        # earth model's basis mat (equivalent to bx <- earth.mod$bx)
bx <- bx[, -1]                       # drop intercept column
bx <- as.data.frame(bx)              # lm requires a data frame
bx$Volume <- trees$Volume            # add Volume to data
lm.mod <- lm(Volume~., data=bx)      # standard linear regression on earth's basis mat
summary(lm.mod)                      # yields same coeffs as above summary
                                     # displayed t values are not meaningful

Ozone readings in Los Angeles with incomplete cases removed

Description

Ozone readings in Los Angeles, with incomplete cases removed.

Format

A data frame with 330 observations on 10 variables.

O3 daily maximum of the hourly average ozone concentrations in Upland, CA
vh 500 millibar pressure height, measured at the Vandenberg air force base
wind wind speed in mph at LAX airport
humidity humidity in percent at LAX
temp Sandburg Air Force Base temperature in degrees Fahrenheit
ibh temperature inversion base height in feet
dpg pressure gradient from LAX to Daggert in mm Hg
ibt inversion base temperature at LAX in degrees Fahrenheit
vis visibility at LAX in miles
doy day of the year

Source

This dataset was copied from library(faraway) and the name changed to ozone1 to prevent a name clash. The data were originally made available by Leo Breiman who was a consultant on a project where the data were generated. Example analyses using these data may be found in Faraway and in Hastie and Tibshirani.

    > ozone1
        O3   vh wind humidity temp  ibh dpg ibt vis doy
    1    3 5710    4       28   40 2693 -25  87 250  33
    2    5 5700    3       37   45  590 -24 128 100  34
    3    5 5760    3       51   54 1450  25 139  60  35
    ...
    330  1 5550    4       85   39 5000   8  44 100 390

References

Faraway (2005) Extending the Linear Model with R https://www.maths.bath.ac.uk/~jjf23

Hastie and Tibshirani (1990) Generalized Additive Models https://hastie.su.domains/pub.htm

See Also

earth

airquality a different set of ozone data


Plot an earth object

Description

Plot an earth object. By default the plot shows model selection, cumulative distribution of the residuals, residuals versus fitted values, and the residual QQ plot.

This function calls plotres internally. The first arguments are identical to plotres.

Usage

## S3 method for class 'earth'
plot(x = stop("no 'x' argument"),

  # the following are identical to plotres arguments

  which = 1:4, info = FALSE, versus = 1, standardize = FALSE, delever = FALSE,
  level = 0, id.n = 3, labels.id = NULL, smooth.col = 2, grid.col = 0,
  jitter = 0, do.par = NULL, caption = NULL,
  trace = 0, npoints = 3000, center = TRUE, type = NULL, nresponse = NA,

  # the following are earth specific

  col.cv = "lightblue", col.grsq = 1, col.rsq = 2, col.infold.rsq = 0,
  col.mean.infold.rsq = 0, col.mean.oof.rsq = "palevioletred",
  col.npreds = if(is.null(object$cv.oof.rsq.tab)) 1 else 0, col.oof.labs = 0,
  col.oof.rsq = "mistyrose2", col.oof.vline = col.mean.oof.rsq,
  col.pch.cv.rsq = 0, col.pch.max.oof.rsq = 0, col.vline = col.grsq,
  col.vseg = 0, lty.grsq = 1, lty.npreds = 2, lty.rsq = 5, lty.vline = "12",
  legend.pos = NULL, ...)

earth_plotmodsel( # for internal use by plotres
  x, col.rsq = 2, col.grsq = 1, col.infold.rsq = 0,
  col.mean.infold.rsq = 0, col.mean.oof.rsq = "palevioletred",
  col.npreds = NULL, col.oof.labs = 0, col.oof.rsq = "mistyrose2",
  col.oof.vline = col.mean.oof.rsq, col.pch.cv.rsq = 0,
  col.pch.max.oof.rsq = 0, col.vline = col.grsq, col.vseg = 0,
  lty.grsq = 1, lty.npreds = 2, lty.rsq = 5, lty.vline = "12",
  legend.pos=NULL, add = FALSE, jitter = 0,
  max.nterms = length(object$rss.per.subset),
  max.npreds=max(1,get.nused.preds.per.subset(object$dirs,object$prune.terms)),
  ...)

Arguments

x

An earth object. This is the only required argument. (It is called "x" for consistency with the generic plot.)

which, info, versus

These arguments are identical to plotres. Please see the help page for plotres.

standardize, delever, level

.

id.n, labels.id, smooth.col

.

grid.col, jitter

.

do.par, caption, trace

.

npoints, center

.

type, nresponse

.

col.cv

Default "lightblue". Color of cross validation line in the residuals plot. This is the residual of the mean out-fold-predicted value.

The following arguments are for the model selection plot.

col.grsq

Default 1. Color of GRSq line in the Model Selection plot. Use 0 for no GRSq line.

col.rsq

Default 2. Color of the RSq line in the Model Selection plot. Use 0 for no RSq line.

col.infold.rsq

Color of in-fold RSq lines for each fold in the Model Selection plot. Applies only if nfold and keepxy were used in the original call to earth. Default is 0, lines not plotted.

col.mean.infold.rsq

Color of mean in-fold RSq for each number of terms in the Model Selection plot. Default is 0, line not plotted. Applies only if nfold and keepxy were used in the original call to earth.

col.mean.oof.rsq

Default "palevioletred". Color of mean out-of-fold RSq for each number of terms in the Model Selection plot. Applies only if nfold and keepxy were used in the original call to earth. Use 0 to not plot this line.

col.npreds

Color of the "number of predictors" plot in the Model Selection plot. The default displays the number of predictors unless the oof.rsq's are displayed. Use 0 for no "number of predictors" plot.

col.oof.labs

Color of fold number labels on the oof.rsq lines. Default is 0, no labels.

col.oof.rsq

Color of out-of-fold RSq lines for each fold in the Model Selection plot. Applies only if nfold and keepxy were used in the original call to earth. Default is "mistyrose2", a pale pink. Use 0 to not plot these lines. May be a vector of colors, which will be recycled if necessary.

col.oof.vline

Color of vertical line at the maximum oof.rsq in the Model Selection plot. Default is col.mean.oof.rsq.

col.pch.cv.rsq

Color of point plotted on the oof.rsq line to indicate the cv.rsq. for that fold (i.e., it is plotted at the number of terms selected by the in-fold GCV). Default is 0, point not plotted.

col.pch.max.oof.rsq

Color of point plotted on the oof.rsq line to indicate the maximum oof.rsq for that fold. Default is 0, point not plotted.

col.vline

Color of the vertical line at selected model in the Model Selection plot. Default is col.grsq. This will be at the maximum GRSq unless pmethod="none". Use 0 for no vertical line.

col.vseg

Default is 0. Color of triangular marker at top of vertical line for best GRSq.

lty.grsq

Line type of GRSq line in the Model Selection plot. Default is 1

lty.npreds

Line type of the "number of predictors" plot in the Model Selection plot. Default is 2.

lty.rsq

Line type of RSq line in the Model Selection plot. Default is 5.

lty.vline

Line type of vertical line at selected model in the Model Selection plot. Default is "12".

legend.pos

Position of the legend in the Model Selection plot. Default is NULL meaning automatic. Use legend.pos=NA or 0 for no legend. Can be something like legend.pos="topleft" or legend.pos=c(6, .75).

add, max.nterms, max.npreds

earth_plotmodsel arguments for internal use by plotres.


...

Please see plotres for the details on the dots arguments.

The ylim argument is treated specially in the model selection plot: ymin equal to -1 means use the smallest GRSq or RSq value, excluding the intercept, and ymax equal -1 means use the largest GRSq or RSq value.

Note

For details on interpreting the graphs, please see the earth package vignettes “Notes on the earth package” and “Variance models in earth”.

Note that cross-validation data will not be displayed unless both nfold and keepxy were used in the original call to earth.

To remove the Number of used predictors from the Model Selection graph (to reduce clutter), use col.npreds=0.

earth_plotmodsel is provided for use by plotres.

Acknowledgment

This function incorporates the function spread.labs from the orphaned package TeachingDemos written by Greg Snow.

See Also

earth, plot.earth.models, plotd, plotmo

Examples

data(ozone1)
earth.mod <- earth(O3 ~ ., data = ozone1, degree = 2)
plot(earth.mod)

Compare earth models by plotting them.

Description

Compare earth models by plotting them.

Usage

## S3 method for class 'earth.models'
plot(x = stop("no 'x' argument"), which = c(1:2),
    caption = "", jitter = 0,
    col.grsq = discrete.plot.cols(length(objects)), lty.grsq = 1,
    col.rsq = 0, lty.rsq = 5,
    col.vline = col.grsq, lty.vline = "12",
    col.npreds = 0,  lty.npreds  = 2,
    legend.text = NULL, do.par = NULL, trace = 0,
    ...)

Arguments

x

A list of one or more earth objects, or a single earth object. This is the only required argument. (This argument is called 'x' for consistency with the generic plot.)

which

Which plots to plot: 1 model, 2 cumulative distribution of residuals. Default is 1:2, meaning both.

caption

Overall caption. Values:
"string" string
"" (default) no caption
NULL generate a caption from the $call component of the earth objects.

jitter

Jitter applied to GRSq and RSq values to minimize over-plotting. Default is 0, meaning no jitter. A typical useful value is 0.01.

For the col arguments below, 0 means do not plot the corresponding graph element. You can use vectors of colors.

col.grsq

Vector of colors for the GRSq plot. The default is discrete.plot.cols(length(x)) which is vector of distinguishable colors, the first three of which are also distinguishable on a monochrome printer. You can examine the colors using
earth:::discrete.plot.cols().

lty.grsq

Line type for the GRSq plot. Default is 1.

col.rsq

Vector of colors for the RSq plot. Default is 0, meaning no RSq plot.

lty.rsq

Line type for the RSq plot. Default is 5.

col.vline

A vertical line is drawn for each object to show which model size was chosen for that object. The color of the line is col.vline. Default is col.grsq.

lty.vline

Line type of vertical lines (a vertical line is drawn to show the selected model for each object). Can be a vector. Default is 3.

col.npreds

Vector of colors for the "number of predictors" plot within the model selection plot. Default is 0, meaning no "number of predictors" plot. The special value NULL means borrow col.grsq (or col.rsq if col.grsq is NULL).

lty.npreds

Line type of the "number of predictors" plot (in the Model Selection plot). Default is 2.

legend.text, do.par, trace

Please see plotres

...

Please see plotres

Note

This function ignores GLM and cross-validation components of the earth model, if any.

See Also

earth, plot.earth, plot.earth.models, plotd, plotmo

Examples

data(ozone1)
a1 <- earth(O3 ~ .,          data = ozone1, degree = 2)
a2 <- earth(O3 ~ .-wind,     data = ozone1, degree = 2)
a3 <- earth(O3 ~ .-humidity, data = ozone1, degree = 2)
plot.earth.models(list(a1,a2,a3), ylim=c(.65,.85))

Plot an evimp object (created by the evimp function)

Description

Plot an evimp object.

Usage

## S3 method for class 'evimp'
plot(x = stop("no 'x' argument"),
    cex.var = 1,
    type.nsubsets = "l", col.nsubsets = "black", lty.nsubsets = 1,
    type.gcv = "l", col.gcv = 2, lty.gcv = 1,
    type.rss = "l", col.rss = "gray60", lty.rss = 1,
    cex.legend = 1, x.legend = nrow(x), y.legend = x[1,"nsubsets"],
    rh.col = 1, do.par = TRUE, ...)

Arguments

x

An evimp object.

cex.var

cex for variable names. Default is 1. Make smaller (say 0.8) if you have lots of variables.

type.nsubsets

Plot type for nsubsets graph. Default is "l". Use "n" for none, "b" looks good too.

col.nsubsets

Color of nsubsets line. Default is "black".

lty.nsubsets

Line type of nsubsets line. Default is 1.

type.gcv, col.gcv, lty.gcv

As above but for the gcv plot

type.rss, col.rss, lty.rss

As above but for the rss plot

cex.legend

cex for legend strings. Default is 1. Make smaller (say 0.8) if you want a smaller legend.

x.legend

x position of legend. Use 0 for no legend.

y.legend

y position of legend.

rh.col

Color of right hand axis label. Use rh.col=0 for no label, a workaround for when the label is mispositioned.

do.par

Call par() for global settings as appropriate. Default is TRUE, which sets
oma=c(bottom.margin,0,0,3), cex=cex.var.
Set to FALSE if you want to append figures to an existing plot.

...

Extra arguments passed to plotting functions.

See Also

earth, evimp, plot.earth.models, plotmo

Examples

data(ozone1)
earth.mod <- earth(O3 ~ ., data=ozone1, degree=2)
ev <- evimp(earth.mod)
plot(ev)
print(ev)

Plot a varmod object (created by calling earth with the varmod argument)

Description

Plot a variance model (a varmod object).

Typically you call this function for a variance model embedded in an earth model.

Usage

## S3 method for class 'varmod'
plot(x = stop("no 'x' argument"), which = 1:4,
  do.par = NULL, info=FALSE,
  cex = NULL, caption = NULL,
  line.col = 2, min.sd.col = line.col,
  trace = 0, ...)

Arguments

x

A varmod object. Typically this is embedded in a parent earth object, and so you invoke this function with plot(earth.mod$varmod). The varmod.method argument must have been specified when building the earth model.

which

Which plots to plot. Default is 1:4 meaning all. The term parent below refers to the earth model in which the varmod is embedded.
1) fitted vs parent fitted
2) fitted vs parent first predictor
3) residuals vs fitted
4) model selection graph (only when varmod.method="earth" or "x.earth").

do.par

Please see plotres

info

Plot some additional information, including lowess fits in the first two plots.

cex

Character expansion.

caption

Default is NULL, meaning automatically generate an overall caption.

line.col

Color of lines in the plots. Default is red.

min.sd.col

Color of the min.sd dotted horizontal line. Default is line.col. Use 0 to not plot this line.

trace, ...

Similar to plotres

Note

The horizontal red dotted line in the first two plots shows the value of min.sd. See earth's varmod.clamp argument.

See Also

varmod

Examples

data(ozone1)

set.seed(1) # optional, for cross validation reproducibility

# note: should really use ncross=30 below but for a quick demo we don't

earth.mod <- earth(O3~temp, data=ozone1, nfold=10, ncross=3, varmod.method="lm")

plot(earth.mod$varmod) # plot the embedded variance model (this calls plot.varmod)

Plot the distribution of predictions for each class

Description

Plot the distribution of the predicted values for each class. Can be used for earth models, but also for models built by lm, glm, lda, etc.

Usage

plotd(object, hist = FALSE, type = NULL, nresponse = NULL, dichot = FALSE,
      trace = FALSE, xlim = NULL, ylim = NULL, jitter = FALSE, main=NULL,
      xlab = "Predicted Value", ylab = if(hist) "Count" else "Density",
      lty = 1, col = c("gray70", 1, "lightblue", "brown", "pink", 2, 3, 4),
      fill = if(hist) col[1] else 0,
      breaks = "Sturges", labels = FALSE,
      kernel = "gaussian", adjust = 1, zero.line = FALSE,
      legend = TRUE, legend.names = NULL, legend.pos = NULL,
      cex.legend = .8, legend.bg = "white", legend.extra = FALSE,
      vline.col = 0, vline.thresh = .5, vline.lty = 1, vline.lwd = 1,
      err.thresh = vline.thresh, err.col = 0, err.border = 0, err.lwd = 1,
      xaxt = "s", yaxt = "s", xaxis.cex = 1, sd.thresh = 0.01, ...)

Arguments

To start off, look at the arguments object, hist, type.
For predict methods with multiple column responses, see the nresponse argument.
For factor responses with more than two levels, see the dichot argument.

object

Model object. Typically a model which predicts a class or a class discriminant.

hist

FALSE (default) to call density internally.
TRUE to call hist internally.

type

Type parameter passed to predict. For allowed values see the predict method for your object (such as predict.earth). By default, plotd tries to automatically select a suitable value for the model in question. (This is "response" for all objects except rpart models, where "vector" is used. The choices will often be inappropriate.) Typically you would set hist=TRUE when type="class".

nresponse

Which column to use when predict returns multiple columns. This can be a column index or column name (which may be abbreviated, partial matching is used). The default is NULL, meaning use all columns of the predicted response.

dichot

Dichotimise the predicted response. This argument is ignored except for models where the observed response is a factor with more than two levels and the predicted response is a numeric vector. The default FALSE separates the response into a group for each factor. With dichot=TRUE the response is separated into just two groups: the first level of the factor versus the remaining levels.

trace

Default FALSE. Use TRUE or 1 to trace plotd — useful to see how plotd partitions the predicted response into classes. Use 2 for more details.

xlim

Limits of the x axis. The default NULL means determine these limits automatically, else specify c(xmin,xmax).

ylim

Limits of the y axis. The default NULL means determine these limits automatically, else specify c(ymin,ymax).

jitter

Jitter the histograms or densities horizontally to minimize overplotting. Default FALSE. Specify TRUE to automatically calculate the jitter, else specify a numeric jitter value.

main

Main title. Values:
"string" string
"" no title
NULL (default) generate a title from the call.

xlab

x axis label. Default is "Predicted Value".

ylab

y axis label. Default is if(hist) "Count" else "Density".

lty

Per class line types for the plotted lines. Default is 1 (which gets recycled for all lines).

col

Per class line colors. The first few colors of the default are intended to be easily distinguishable on both color displays and monochrome printers.

fill

Fill color for the plot for the first class. For hist=FALSE, the default is 0, i.e., no fill. For hist=TRUE, the default is the first element in the col argument.

breaks

Passed to hist. Only used if hist=TRUE. Default is "Sturges". When type="class", setting breaks to a low number can be used to widen the histogram bars

labels

TRUE to draw counts on the hist plot. Only used if hist=TRUE. Default is FALSE.

kernel

Passed to density. Only used if hist=FALSE. Default is "gaussian".

adjust

Passed to density. Only used if hist=FALSE. Default is 1.

zero.line

Passed to plot.density. Only used if hist=FALSE. Default is FALSE.

legend

TRUE (default) to draw a legend, else FALSE.

legend.names

Class names in legend. The default NULL means determine these automatically.

legend.pos

Position of the legend. The default NULL means position the legend automatically, else specify c(x,y).

cex.legend

cex for legend. Default is .8.

legend.bg

bg color for legend. Default is "white".

legend.extra

Show (in the legend) the number of occurrences of each class. Default is FALSE.

vline.thresh

Horizontal position of optional vertical line. Default is 0.5. The vertical line is intended to indicate class separation. If you use this, don't forget to set vline.col.

vline.col

Color of vertical line. Default is 0, meaning no vertical line.

vline.lty

Line type of vertical line. Default is 1.

vline.lwd

Line width of vertical line. Default is 1.

err.thresh

x axis value specifying the error shading threshold. See err.col. Default is vline.thresh.

err.col

Specify up to three colors to shade the "error areas" of the density plot. The default is 0, meaning no error shading. This argument is ignored unless hist=FALSE. If there are more than two classes, err.col uses only the first two. This argument is best explained by running an example:

data(etitanic)
earth.mod <- earth(survived ~ ., data=etitanic)
plotd(earth.mod, vline.col=1, err.col=c(2,3,4))
      

The three areas are (i) the error area to the left of the threshold, (ii) the error area to the right of the threshold, and, (iii) the reducible error area. If less than three values are specified, plotd re-uses values in a sensible manner. Use values of 0 to skip areas. Disjoint regions are not handled well by the current implementation.

err.border

Borders around the error shading. Default is 0, meaning no borders, else specify up to three colors.

err.lwd

Line widths of borders of the error shading. Default is 1, else specify up to three line widths.

xaxt

Default is "s". Use xaxt="n" for no x axis.

yaxt

Default is "s". Use yaxt="n" for no y axis.

xaxis.cex

Only used if hist=TRUE and type="class". Specify size of class labels drawn on the x axis. Default is 1.

sd.thresh

Minimum acceptable standard deviation for a density. Default is 0.01. Densities with a standard deviation less than sd.thresh will not be plotted (a warning will be issued and the legend will say "not plotted").

...

Extra arguments passed to the predict method for the object.

Note

This function calls predict with the data originally used to build the model, and with the type specified above. It then separates the predicted values into classes, where the class for each predicted value is determined by the class of the observed response. Finally, it calls density (or hist if hist=TRUE) for each class-specific set of values, and plots the results.

This function estimates distributions with the density and hist functions, and also calls plot.density and plot.histogram. For an overview see Venables and Ripley MASS section 5.6.

Partitioning the response into classes

Considerable effort is made to partition the predicted response into classes in a sensible way. This is not always possible for multiple column responses and the nresponse argument should be used where necessary. The partitioning details depend on the types and numbers of columns in the observed and predicted responses. These in turn depend on the model object and the type argument.

Use the trace argument to see how plotd partitions the response for your model.

Degenerate densities

A message such as
Warning: standard deviation of "male" density is 0, density is degenerate?
means that the density for that class will not be plotted (the legend will say "not plotted").

Set sd.thresh=0 to get rid of this check, but be aware that histograms (and sometimes x axis labels) for degenerate densities will be misleading.

Using plotd for various models

This function is included in the earth package but can also be used with other models.

Example with glm:

      library(earth); data(etitanic)
      glm.model <- glm(sex ~ ., data=etitanic, family=binomial)
      plotd(glm.model)
  

Example with lm:

      library(earth); data(etitanic)
      lm.model <- lm(as.numeric(sex) ~ ., data=etitanic)
      plotd(lm.model)
  

Using plotd with lda or qda

The plotd function has special handling for lda (and qda) objects. For such objects, the type argument can take one of the following values:

"response" (default) linear discriminant
"ld" same as "response"
"class" predicted classes
"posterior" posterior probabilities

Example:

    library(MASS); library(earth); data(etitanic)
    lda.model <- lda(sex ~ ., data=etitanic)
    plotd(lda.model) # linear discriminant by default
    plotd(lda.model, type="class", hist=TRUE, labels=TRUE)
    

This handling of type is handled internally by plotd and type is not passed to predict.lda (type is used merely to select fields in the list returned by predict.lda). The type names can be abbreviated down to a single character.

For objects created with lda.matrix (as opposed to lda.formula), plotd blindly assumes that the grouping argument was the second argument.

plotd does not yet support objects created with lda.data.frame.

For lda responses with more than two factor levels, use the nresponse argument to select a column in the predicted response. Thus with the default type=NULL, (which gets automatically converted by plotd to type="response"), use nresponse=1 to select just the first linear discriminant. The default nresponse=NULL selects all columns, which is typically not what you want for lda models. Example:

    library(MASS); library(earth);
    set.seed(1)      # optional, for reproducibility
    example(lda)     # creates a model called "z"
    plot(z, dimen=1) # invokes plot.lda from the MASS package
    plotd(z, nresponse=1, hist=1) # equivalent using plotd
                                 # nresponse=1 selects first linear discr.
    

The dichot=TRUE argument is also useful for lda responses with more than two factor levels.

TODO

Handle degenerate densities in a more useful way.
Add freq argument for hist.

See Also

density, plot.density
hist, plot.histogram
earth, plot.earth

Examples

if (require(earth)) {
    old.par <- par(no.readonly=TRUE);
    par(mfrow=c(2,2), mar=c(4, 3, 1.7, 0.5), mgp=c(1.6, 0.6, 0), cex = 0.8)
    data(etitanic)
    mod <- earth(survived ~ ., data=etitanic, degree=2, glm=list(family=binomial))

    plotd(mod)

    plotd(mod, hist=TRUE, legend.pos=c(.25,220))

    plotd(mod, hist=TRUE, type="class", labels=TRUE, xlab="", xaxis.cex=.8)

    par(old.par)
}

Predict with an earth model

Description

Predict with an earth model.

Usage

## S3 method for class 'earth'
predict(object = stop("no 'object' argument"), newdata = NULL,
        type = c("link", "response", "earth", "class", "terms"),
        interval = "none", level = .95,
        thresh = .5, trace = FALSE, ...)

Arguments

object

An earth object. This is the only required argument.

newdata

Make predictions using newdata, which can be a data frame, a matrix, or a vector with length equal to a multiple of the number of columns of the original input matrix x.
Default is NULL, meaning return values predicted from the training set.
NAs are allowed in newdata (and the predicted value will be NA unless the NAs are in variables that are unused in the earth model).

type

Type of prediction. One of "link" (default), "response", "earth", "class", or "terms". See the Note below.

interval

Return prediction or confidence levels. Default is "none". Use interval="pint" to get prediction intervals on new data.
Requires that the earth model was built with varmod.method.
This argument gets passed on as the type argument to predict.varmod. See its help page for details.

level

Confidence level for the interval argument. Default is 0.95, meaning construct 95% confidence bands (estimate the 2.5% and 97.5% levels).

thresh

Threshold, a value between 0 and 1 when predicting a probability. Only applies when type="class". Default is 0.5. See the Note below.

trace

Default FALSE. Set to TRUE to see which data, subset, etc. predict.earth is using.

...

Unused, but provided for generic/method consistency.

Value

The predicted values (a matrix for multiple response models).

If type="terms", a matrix with each column showing the contribution of a predictor.

If interval="pint" or "cint", a matrix with three columns:
fit: the predicted values
lwr: the lower confidence or prediction limit
upr: the upper confidence or prediction limit

If interval="se", the standard errors.

Note

Predicting with standard earth models

Use the default type="link", or possibly type="class".

Actually, the "link", "response", and "earth" choices all return the same value unless the glm argument was used in the original call to earth.

Predicting with earth-GLM models

This section applies to earth models with a GLM component, i.e., when the glm argument was used in the original call to earth.

The "link" and "response" options: see predict.glm for a description of these. In brief: for logistic models use type="response" to get probabilities, and type="link" to get log-odds.

Use option "earth" to get the linear fit (this gives the prediction you would get if your original call to earth had no glm argument).

Predicting with "class"

Use option "class" to get the predicted class. With option "class", this function first makes predictions with type="response" and then assigns the predicted values to classes as follows:

(i) When the response is a logical, predict TRUE if the predicted probability is greater than thresh (default 0.5).

(ii) When the response is a numeric, predict TRUE if the predicted value is greater than thresh. Actually, this is identical to the above case, although thresh here may legitimately be a value outside the 0...1 range.

(iii) When the response is a two level factor, predict the second level if its probability is more than thresh. In other words, with the default thresh=0.5 predict the most probable level.

(iv) When the response is a three or more level factor, predict the most probable level (and thresh is ignored).

Predicting with "terms"

The "terms" option returns a "link" response suitable for termplot. Only the additive terms and the first response (for multi-response models) are returned. Also, "terms" always returns the earth terms, and ignores the GLM component of the model, if any.

See Also

earth, predict

Examples

data(trees)
earth.mod <- earth(Volume ~ ., data = trees)
predict(earth.mod) # same as earth.mod$fitted.values
predict(earth.mod, data.frame(Girth=10, Height=80)) # yields 17.6
predict(earth.mod, c(10,80))                        # equivalent

Predict with a varmod model

Description

You probably won't need to call this function directly. It is called by predict.earth when that function's interval argument is used.

Usage

## S3 method for class 'varmod'
predict(
    object  = stop("no 'object' argument"),
    newdata = NULL,
    type    = c("pint", "cint", "se", "abs.residual"),
    level   = .95,
    trace   = FALSE,
    ...)

Arguments

object

A varmod object.

newdata

Make predictions using newdata. Default is NULL, meaning return values predicted from the training set.

type

Type of prediction. This is the interval argument of predict.earth. One of

"pint" Prediction intervals.

"cint" Confidence intervals. Cannot be used with newdata.

"se" Standard error of the parent model residuals.

"abs.residual" The absolute residuals of the parent model on which the residual model regresses.

level

Confidence level for the interval argument. Default is .95, meaning construct 95% confidence bands (estimate the 2.5% and 97.5% levels).

trace

Currently unused.

...

Unused, but provided for generic/method consistency.

Note

predict.varmod is called by predict.earth when its interval argument is used.

See Also

predict.earth varmod

Examples

data(ozone1)

set.seed(1) # optional, for cross validation reproducibility

# note: should really use ncross=30 below but for a quick demo we don't

earth.mod <- earth(O3~temp, data=ozone1, nfold=10, ncross=3, varmod.method="lm")

# call predict.earth, which calls predict.varmod

predict(earth.mod, newdata=ozone1[200:203,], interval="pint", level=.95)

Residuals for an earth model

Description

Residuals of an earth model.

Usage

## S3 method for class 'earth'
residuals(object = stop("no 'object' argument"),
          type = NULL, warn = TRUE, ...)

Arguments

object

An earth object. This is the only required argument.

type

One of:

"earth" (default) Residuals from the lm fit on bx.
"response" Residuals as above, but for earth-glm models return the glm response residuals.
"standardize" Residuals divided by se * sqrt(1 - h_ii). See the standardize argument of plot.earth.
"delever" Residuals divided by sqrt(1 - h_ii). See the delever argument of plot.earth.

The following options are for earth-glm models. They return the GLM residuals (from the glm fit on bx). See residuals.glm for details:

"deviance"
"pearson"
"working"
"partial"

The following options for earth-glm models are redundant. They are provided for compatibility with older versions of earth or other functions:

"glm.response" same as "response"
"glm.deviance" same as "deviance"
"glm.pearson" same as "pearson"
"glm.working" same as "working"
"glm.partial" same as "partial"

warn

This function gives warnings when the results are not what you may expect. Use warn=FALSE to turn of just these warnings.

...

Unused, but provided for generic/method consistency.

Value

The residual values (will be a matrix for multiple response models).

See Also

earth
residuals
resid identical to residuals

Examples

data(etitanic)
earth.mod <- earth(pclass ~ ., data=etitanic, glm=list(family=binomial))
head(resid(earth.mod, warn=FALSE))      # earth residuals, a column for each response
head(resid(earth.mod, type="response")) # GLM response resids, a column for each response

Summary method for earth objects

Description

Summary method for earth objects.

Usage

## S3 method for class 'earth'
summary(object = stop("no 'object' argument"),
        details = FALSE,  style = c("h", "pmax", "max", "C", "bf"),
        decomp = "anova", digits = getOption("digits"), fixed.point=TRUE,
        newdata = NULL, ...)

## S3 method for class 'summary.earth'
print(x = stop("no 'x' argument"),
        details = x$details,
        decomp = x$decomp, digits = x$digits, fixed.point = x$fixed.point,
        newdata = x$newdata, ...)

Arguments

object

An earth object. This is the only required argument for summary.earth.

x

A summary.earth object. This is the only required argument for print.summary.earth.

details

Default is FALSE. Use TRUE to print more information about earthglm models. But note that the displayed Standard Errors and statistics for the GLM coefficients are meaningless (because of the amount of preprocessing done by earth to select the regression terms).

style

Formatting style. One of
"h" (default) more compact
"pmax" for those who prefer it and for compatibility with old versions of earth
"max" is the same as "pmax" but prints max rather than pmax
"C" C style expression with zero based indexing
"bf" basis function format.

decomp

Specify how terms are ordered. Default is "anova". Use "none" to order the terms as created by the forward.pass. See format.earth for a full description.

digits

The number of significant digits.
For summary.earth, the default is getOption("digits").
For print.summary.earth, the default is the $digits component of object.

fixed.point

Method of printing numbers in matrices. Default is TRUE which prints like this (making it easier to compare coefficients):

        (Intercept)    15.029
        h(temp-58)      0.313
        h(234-ibt)     -0.046
        ...

whereas fixed.point=FALSE prints like this (which is more usual in R):

        (Intercept)   1.5e+01
        h(temp-58)    3.1e-01
        h(234-ibt)   -4.6e-02
        ...

Matrices with two or fewer rows are never printed with a fixed point.

newdata

Default NULL.
Else print R-Squared for the new data (and the returned object will have newrsq and newdata fields). Additionally, if a variance model is present print the interval coverage table for the new data.

...

Extra arguments are passed to format.earth.

Value

The value is the same as that returned by earth but with the following extra components.

strings

String(s) created by format.earth. For multiple response models, a vector of strings.

newrsq

Only if newdata was passed to summary.earth.

newdata

Only if newdata was passed to summary.earth.

digits
details
decomp
fixed.point

The corresponding arguments, passed on to print.summary.earth.

Note

The printed Estimated importance uses evimp with the nsubsets criterion. The most important predictor is printed first, and so on.

See Also

earth, evimp, format.earth

Examples

earth.mod <- earth(Volume~ ., data = trees)
summary(earth.mod, digits = 2)

Update an earth model

Description

Update an earth model.

Usage

## S3 method for class 'earth'
update(object = stop("no 'object' argument"),
       formula. = NULL, ponly = FALSE, ..., evaluate = TRUE)

Arguments

object

The earth object

formula.

The formula. argument is treated like earth's formula argument.

ponly

Force pruning only, no forward pass. Default is FALSE, meaning update.earth decides automatically if a forward pass is needed. See note below.

...

Arguments passed on to earth.

evaluate

If TRUE (default) evaluate the new call, else return the call. Mostly for compatibility with the generic update.

Details

If only the following arguments are used, a forward pass is unnecessary, and update.earth will perform only the pruning pass. This is usually much faster for large models.

     object
     glm
     trace
     nprune
     pmethod
     Eval.model.subsets
     Print.pruning.pass
     Force.xtx.prune
     Use.beta.cache
     Endspan.penalty
     Get.leverages

This automatic determination to do a forward pass can be overridden with the ponly argument. If ponly=TRUE the forward pass will be skipped and only the pruning pass will be executed. This is useful for doing a pruning pass with new data. (Use earth's data argument to specify the new data.) Typically in this scenario you would also specify penalty=-1. This is because with sufficient new data, independent of the original training data, the RSS not the GCV should be used for evaluating model subsets (The GCV approximates what the RSS would be on new data — but here we actually have new data, so why bother approximating. This "use new data for pruning" approach is useful in situations where you don't trust the GCV approximation for your data.) By making penalty=-1, earth will calculate the RSS, not the GCV. See also the description of penalty on the earth help page.

Another (somewhat esoteric) use of ponly=TRUE is to do subset selection with a different penalty from that used to build the original model.

With trace=1, update.earth will tell you if earth's forward pass was skipped.

If you used keepxy=TRUE in your original call to earth, then update.earth will use the saved values of x, y, etc., unless you specify otherwise by arguments to update.earth. It can be helpful to set trace=1 to see which x and y is used by update.earth.

Value

The value is the same as that returned by earth. If object is the only parameter then no changes are made — the returned value will be the same as the original object.

See Also

earth

Examples

data(ozone1)

(earth.mod <- earth(O3 ~ ., data = ozone1, degree = 2))

update(earth.mod, formula = O3 ~ . - temp) # requires forward pass and pruning

update(earth.mod, nprune = 8)              # requires only pruning

update(earth.mod, penalty=1, ponly=TRUE)   # pruning pass only with a new penalty

Variance models for estimating prediction intervals

Description

A variance model estimates the variance of predicted values. It can be used to estimate prediction intervals. See the interval argument of predict.earth.

A variance model is built by earth if earth's varmod.method argument is specified. Results are stored in the $varmod field of the earth model. See the vignette “Variance models in earth” for details.

You probably won't need to directly call print.varmod or summary.varmod. They get called internally by summary.earth.

Usage

## S3 method for class 'varmod'
summary(
    object  = stop("no 'object' argument"),
    level   = .95,
    style   = "standard",
    digits  = 2,
    newdata = NULL,
    ...)

Arguments

object

A varmod object. This is the only required argument.

level

Same as predict.earth's level argument.

style

Determines how the coefficients of the varmod are printed by summary.varmod:
"standard" (default)
"unit" for easy comparison normalize the coefficients by dividing by the first coefficient.

digits

Number of digits to print. Default is 2.

newdata

Default NULL.
Else print the interval coverage table for the new data.

...

Dots are passed on.

Note

A "varmod" object has the following fields:

  • call The call used internally in the parent model to build the varmod object.

  • parent The parent earth model.

  • method Copy of the varmod.method argument to the parent model.

  • package NULL, unless method="gam", in which case either "gam" or "mgcv".

  • exponent Copy of the varmod.exponent argument to the parent model.

  • lambda Currently always 1, meaning use absolute residuals.

  • rmethod Currently always "hc2", meaning correct the residuals with 1/(1-h_ii).

  • converged Did the residual submodel IRLS converge?

  • iters Number of residual model IRLS iterations (1 to 50).

  • residmod The residual submodel. So for example, if varmod.method="lm", this will be an lm object.

  • min.sd The predicted residual standard deviation is clamped so it will always be at least this value. This prevents prediction of negative or absurdly small variances. See earth's varmod.clamp argument. Clamping takes place in predict.varmod, which is called by predict.earth when estimating prediction intervals.

  • model.var An n x 1 matrix. The model.var for an observation is the estimated model variance for that observation over all datasets, and is estimated with repeated cross validation. It is the variance of the mean out-of-fold prediction for that observation over ncross repetitions.

  • abs.resids An n x 1 matrix. The absolute residuals used to build the residual model.

  • parent.x An n x p matrix. Parent earth model x.

  • parent.y An n x 1 matrix. Parent earth model y.

  • iter.rsq Weighted R-Squared of residual submodel residmod, after IRLS iteration.

  • iter.stderr Standard errors of the coefficients of the residual submodel residmod, after IRLS iteration.

See Also

plot.varmod, predict.varmod

Examples

data(ozone1)

set.seed(1) # optional, for cross validation reproducibility

# note: should really use ncross=30 below but for a quick demo we don't

earth.mod <- earth(O3~temp, data=ozone1, nfold=10, ncross=3, varmod.method="lm")

print(summary(earth.mod)) # note additional info on the variance model

old.mfrow <- par(mfrow=c(2,2), mar=c(3, 3, 3, 1), mgp=c(1.5, 0.5, 0))

plotmo(earth.mod, do.par=FALSE, response.col=1, level=.90, main="earth model: O3~temp")

plot(earth.mod, which=3, level=.90) # residual plot: note 90% pred and darker conf intervals

par(par=old.mfrow)