coefPlot: Plots coefficients in an impulse response format

In expandFunctions: Feature Matrix Builder

Description

Given a model xObj for which coef(xObj) returns a set of coefficients, plot the coefficients.

The plots make it easier to compare which features are large, which are set to zero, and how features change from run to run in a graphical manner.

If the fitting process is linear (e.g. lm, glmnet, etc.) and the original features are appropriately ordered lags, this will generate an impulse response.

Any coefficients that are exactly zero (for instance, set that way by LASSO) will appear as red X's; non-zero points will be black O's.

Usage

coefPlot(xObj, includeIntercept = FALSE, type = "h", main = NULL, ...)

Arguments

xObj	Output of a fitting model.
includeIntercept	Should the 1st coefficient be plotted? Default is FALSE.
type	Graphics type. Default is "h", which results in an impulse-like plot.
main	"main" title; default is the relative number of non-zero coefficients, a measure of sparsity.
...	Optional additional graphical parameters, for instance to set ylim to a fixed value.

Details

If includeIntercept==TRUE, the intercept of the model will be plotted as index 0.

Changing the type using type="b" will result in a parallel coordinate-like plot rather than an impulse-like plot. It is sometimes easier to see the differences in coefficients with type="b" rather than type="h".

Value

Invisibly returns TRUE. Used for its graphic side effects only.

Examples

set.seed(1)
nObs <- 100
X <- distMat(nObs,6)
A <- cbind(c(1,0,-1,rep(0,3))) # Y will only depend on X[,1] and X[,3]
Y <- X %*% A + 0.1*rnorm(nObs)
lassoObj <- easyLASSO(X,Y)
Yhat <- predict(lassoObj,newx=X)
yyHatPlot(Y,Yhat)
coef( lassoObj ) # Sparse coefficients
coefPlot( lassoObj )
coefPlot( lassoObj, includeIntercept=TRUE )
coefPlot( lassoObj, type="b" )

Example output

7 x 1 sparse Matrix of class "dgCMatrix"
                     s0
(Intercept) -0.01817725
V1           0.97668965
V2           .         
V3          -0.97447289
V4           .         
V5           .         
V6           .         
Warning messages:
1: In mean.default(!myZeros) :
  argument is not numeric or logical: returning NA
2: In mean.default(xCoef != 0) :
  argument is not numeric or logical: returning NA

expandFunctions documentation built on May 2, 2019, 9:15 a.m.