pplot: Plot Regression Terms using ggplot2

In thomaspruf/fiwidat: Datasets for Vienna VetMedUni Statistics Class

Plot Regression Terms using ggplot2

Description

Plots regression terms against their predictors. Generates ggplots based on library visreg

Usage

pplot=function (model, xvar, by = NULL, breaks = NULL, partial = FALSE,
          type = c("conditional", "contrast"), scale = c("linear",
        "response"), leg.name = NULL, leg.labs = NULL, cols = c("red",
        "blue", "green", "purple"), linesize = 1.5, transp = 0.5,
          rsize = 2, rcolor = "red", rshade = FALSE, rshapes = TRUE)

Arguments

model	A fitted model.
xvar	Against which predictor to plot.
by	Used for interactions. Plot are shown for levels of by. Continuous variables are changed to factors
breaks	If a single number, number of levels. If 2 or more numbers, discrete values/levels of by.
partial	Logical. Should residuals been shown?
type	Use 'contrast' for models with random factors
scale	Linear or response. If scale='response' for a glm, the inverse link function will be applied so that the model is plotted on the scale of the original response.
leg.name	Optional. Name of legend.
leg.labs	Optional. Legend labels.
cols	Colors of confidence bands.
linesize	Number. Set thickness of regression line?
transp	Transparency of cols (0-1).
rsize	Size of symbols for residuals.
rcolor	Color of residuals.
rshade	Logical. Should residuals been shown in different shades of gray per level of by?
rshapes	Logical. Should residuals been shown in different shapeas per level of by?

Details

Wrapper for essentials of visreg. See ?visreg for more.

Author(s)

T. Ruf

References

Breheny P and Burchett W (2017). Visualization of Regression Models Using visreg. The R Journal, 9: 56-71.

Examples

require(visreg)
require(ggplot2)
devAskNewPage(ask =TRUE)

fit <- lm(Ozone ~ Solar.R + Wind * Temp, data=airquality)

theme_set(theme_rufus(base_size=20))

pplot(fit,xvar="Wind",by="Temp", breaks=c(60,90),partial=TRUE,
      leg.name="Temperature",leg.labs=c("cool","warm"))

pplot(fit,xvar="Wind",by="Temp", breaks=c(3),partial=TRUE,
      leg.name="Temperature",leg.labs=c("cool","average","warm"),
      cols=c("yellow","orange","red"),rsize=3,rshapes=TRUE)

pplot(fit,xvar="Wind",by="Temp", breaks=c(3),partial=FALSE,
      leg.name="Temperature",leg.labs=c("cool","average","warm"),
      cols=c("yellow","orange","red"),rsize=3)+
      theme(legend.position = c(0.8, 0.8))

require(mgcv)
require (nlme)

mod.lme=lme(distance~age*Sex, random=~1|Subject, data=Orthodont)
pplot(mod.lme,xvar="age",by="Sex",partial=TRUE,leg.name="Gender",type='contrast')


mod.bam=bam(distance~age*Sex+s(Subject, bs = 're'), data=Orthodont)
pplot(mod.bam,xvar="age",by="Sex",type='contrast')
pplot(mod.bam,xvar="age",by="Sex",partial=TRUE,leg.name="Gender",type='contrast')
# bam can be used to compute LINEAR mixed effects and generates confidence intervals.
# Coefficients, t and p-values are virtually identical to lme and lmer.

# pplot handles simple gams too.
fit <- gam(Ozone ~ s(Solar.R) + s(Wind) + s(Temp), data=airquality)
pplot(fit,xvar="Wind",partial=TRUE)

thomaspruf/fiwidat documentation built on Aug. 28, 2023, 11:12 p.m.