title | description | keywords | author | ms.author | manager | ms.date | ms.topic | ms.service | ms.assetid | ROBOTS | audience | ms.devlang | ms.reviewer | ms.suite | ms.tgt_pltfrm | ms.custom |
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
rxStepPlot function (revoAnalytics) | Microsoft Docs |
Plot stepwise coefficients for rxLinMod, rxLogit and rxGlm objects. |
(revoAnalytics), rxStepPlot, hplot |
chuckheinzelman |
charlhe |
cgronlun |
07/15/2019 |
reference |
mlserver |
Plot stepwise coefficients for rxLinMod
, rxLogit
and rxGlm
objects.
rxStepPlot(x, plotStepBreaks = TRUE, omitZeroCoefs = TRUE, eps = .Machine$double.eps,
main = deparse(substitute(x)), xlab = "Step", ylab = "Stepwise Coefficients",
type = "b", pch = "*",
... )
a stepwise regression object of class rxLinMod
, rxLogit
, or rxGlm
with a non-empty stepCoefs
component, which can be generated by setting the component keepStepCoefs
of the variableSelection
argument to TRUE
when the model is fitted.
logical value. If TRUE
, vertical lines are drawn at each step in the stepwise coefficient paths.
logical flag. If TRUE
, coefficients with sum of absolute values less than eps
will be omitted for plotting.
the smallest positive floating-point number that is treated as nonzero.
additional graphical arguments to be passed directly to the underlying matplot function.
the nonzero stepwise coefficients are returned invisibly.
Microsoft Corporation Microsoft Technical Support
rxStepControl, rxLinMod, rxLogit, rxGlm.
## setup
form <- Sepal.Length ~ Sepal.Width + Petal.Length
scope <- list(
lower = ~ Sepal.Width,
upper = ~ Sepal.Width + Petal.Length + Petal.Width * Species)
## rxLinMod/variableSelection with keepStepCoefs = TRUE
varsel <- rxStepControl(method = "stepwise", scope = scope, keepStepCoefs = TRUE)
rxlm.step <- rxLinMod(form, data = iris, variableSelection = varsel,
verbose = 1, dropMain = FALSE, coefLabelStyle = "R")
rxStepPlot(rxlm.step)