quickPredict {STAR} | R Documentation |
Designed to quickly compute the effect of a single model term. This term can correspond to a single variable effect or to the interaction of two variables.
quickPredict(object, include = object$terms$labels[2], se.fit = TRUE, length.out, otherTermsFct = median) object %qp% include
object |
an object inheriting from ssanova and ssanova0
(gssanova and gssanova0 objects are therefore suitable).
|
include |
a character string corresponding to a single
model term. See predict.ssanova and predict.ssanova .
|
se.fit |
logical flag indicating if standard errors are
required. See predict.ssanova and predict.ssanova .
|
length.out |
a positive integer, the number of points at which
the prediction should be performed. These points are uniformly spread
on the definition domain of the variable(s) implicitely specified by
argument include . If missing a default of 501 for terms
involving a single variable and of 101 for interaction terms involving
two variables is provided.
|
otherTermsFct |
a function applied to the other variables required for model specification. |
%qp%
is the binary version of quickPredict
.
A quickPredict
object. This object is a list
with the following
components:
xx |
a numeric vector with the values of the variable specified by
the model term selected by argument include . When an interaction term was
selected the values of the first variable are stored here. |
yy |
a numeric vector with the values of the second variable specified by
the interaction term selected by argument include . When
selected term is not an interaction term, this component is
NULL . |
include |
the value of the argument with this name. |
call |
the matched call. |
est.mean |
a numeric vector or matrix, for intercation terms, containing the estimated mean of the term. |
est.sd |
a numeric vector or matrix, for intercation terms,
containing the estimated SD of the term. Is NULL is argument
se.fit was FALSE . |
Christophe Pouzat christophe.pouzat@gmail.com
predict.ssanova
,
predict.ssanova
,
plot.quickPredict
,
image.quickPredict
,
contour.quickPredict
,
persp.quickPredict
,
plot.ssanova
## Follow up of ssanova example of gss data(nox) nox.fit <- ssanova(log10(nox)~comp*equi,data=nox) ## get prediction for the first term, comp comp.pred <- quickPredict(nox.fit) ## plot result with method plot for quickPredict objects plot(comp.pred) ## get prediction for the second term, equi using the binary version equi.pred <- nox.fit %qp% "equi" plot(equi.pred) ## get prediction for the interaction term, comp:equi comp.equi.pred <- nox.fit %qp% "comp:equi" ## use image method image image(comp.equi.pred) ## use contour method contour(comp.equi.pred,col=2,lwd=2,labcex=1.5) contour(comp.equi.pred,what="sd",lty=3,labcex=1.2,add=TRUE) ## use persp method persp(comp.equi.pred,theta=-10,phi=20)