margplot: Plotting marginal effects for interactions in nonlinear...

In connorjmccabe/modglm: Computing interaction effects for nonlinear probability and counts

Description

Plotting marginal effects for interactions in nonlinear probability and count models

Usage

margplot(
  model,
  vars,
  data,
  hyps = "means",
  foc,
  mod,
  modlevels,
  modnames = NULL,
  sm = T
)

Arguments

model	The estmiated model object. Currently, this may take logit or Poisson model objects estimated using the 'stats' package or negative binomial model objects tested using the 'MASS" package.
vars	The interacting variable names.
data	The data frame on which the estimated model is based.
hyps	User-specified levels of the predictor variables for displaying marginal effects. By default, this is specified at the mean values of all included covariates.
foc	Name of the variable that will be the focal (i.e., x-axis) variable
mod	Name of hte variable that will be the moderator (i.e., legend) variable
modlevels	A vector of hypothetical levels of the moderator at which to display marginal effects.
modnames	If the user wants to call the moderator levels something other Than what is provided in the data itself, the user can supply a vector of names here.
sm	Stands for "small multiple". If TRUE, marginal effects will be displayed as small multiple (i.e., trellis) plots. if FALSE, all marginal effects will appear on a single plot.

Examples

#Building from the modglm example, this outputs a ggplot object plotting the effect of x1 on
#the count of y across males and females:
set.seed(1678)
library(ggplot2)
b0 <- -3.8 ##Intercept
b1 <- .35 ###Sensation Seeking Effect
b2 <- .9 #Premeditation  Effect
b3 <- 1.1 #Sex covariate effect
b13<- .2 #product term coefficient
n<-1000 #Sample Size
mu<-rep(0,2) #Specify means
S<-matrix(c(1,.5,.5,1),nrow=2,ncol=2) #Specify covariance matrix
sigma <- 1 #Level 1 error

rawvars<-MASS::mvrnorm(n=n, mu=mu, Sigma=S) #simulates our
#continuous predictors from a multivariate normal distribution
cat<-rbinom(n=n,1,.5)
id<-seq(1:n)
eij <- rep(rnorm(id, 0, sigma))
xb<-  (b0) + (b1) * (rawvars[,1]) + (b2) * (rawvars[,2]) + (b3)*cat + b13*cat*(rawvars[,1]) + eij

ct <- exp(xb)
y <- rpois(n,ct)

df <- data.frame(y=y,senseek=rawvars[,1],premed=rawvars[,2],male=cat)

#premeditation, sex, and the interaction between sensation seeking and sex:

pois<-glm(y ~ senseek + premed + male + senseek:male, data=df,family="poisson")

p<-margplot(model=pois, vars=c("senseek","male"),foc="senseek", mod="male",
modlevels=c(0,1),data=df, hyps="means",sm=FALSE,modnames=c("Female","Male"))

#Because this is a ggplot object, we can modify anything further by adding elements to this plot. E.g.:
p + ylab("Count")

connorjmccabe/modglm documentation built on March 28, 2021, 10:45 p.m.