Region Plots for IRT Models
Base graphics plotting function for region plot visualization of IRT models.
regionplot(object, parg = list(type = NULL, ref = NULL, alias = TRUE),
names = TRUE, main = NULL, xlab = "", ylab = "Latent trait", ylim = NULL,
off = 0.1, col = NULL, linecol = 2, srt = 45, adj = c(1.1, 1.1),
axes = TRUE, ...)object |
a fitted model object of class |
parg |
list of arguments passed over to internal calls of
|
names |
logical or character. If |
main |
character, specifying the overall title of the plot. |
xlab, ylab |
character, specifying the x and y axis labels. |
ylim |
numeric, specifying the y axis limits. |
off |
numeric, the distance (in scale units) between two item rectangles. |
col |
character, list or function, specifying the colors of the regions.
Either a single vector with k color names, a list with m
elements and each element is a character vector with color names for the
regions of item j or a color-generating function like, e.g.,
|
linecol |
color for lines indicating “hidden” categories. |
srt, adj |
numeric. Angle ( |
axes |
logical. Should axes be drawn? |
... |
further arguments passed to |
The region plot visualization implemented here was already used by Van der
Linden and Hambleton (1997) in the context of IRT and has been called "effect
plots" by Fox & Hong (2009). In our implementation, these plots show,
dependent on the chosen type of threshold parameters, different regions for
the categories of an item over the theta axis. If type is set to
"modus", the cutpoints correspond to the threshold parameters and the
rectangles mark the theta regions where a category is the single most probable
category chosen with a certain value of the latent trait. If type is
set to "median", the cutpoints correspond to the point on the theta
axis, where the cumulative probability to score in category k or higher
is 0.5, i.e., P(X_{ij} ≥q k) = 0.5. If set to "mean", the
cutpoints correspond to the point on the theta axis where the expected score
E(X_{ij}) is exactly between two categories, e.g., 0.5 for a dichotomous
item.
If type is set to "mode" and there are unordered threshold
parameters, the location of the original threshold parameters are indicated by
red dashed lines.
Fox J, Hong J (2009). Effect Displays in R for Multinomial and Proportional-Odds Logit Models: Extensions to the effects Package. Journal of Statistical Software, 32(1), 1–24.
Van der Linden WJ, Hambleton RK (1997). Handbook of Modern Item Response Theory. Springer, New York.
## load verbal aggression data
data("VerbalAggression", package = "psychotools")
## fit a Partial credit model to the items of the first other-to-blame
## situation: "A bus fails to stop for me"
pcm <- pcmodel(VerbalAggression$resp[, 1:6])
## a region plot with modus as cutpoint and custom labels
lab <- paste(rep(c("Curse", "Scold", "Shout"), each = 2),
rep(c("Want", "Do"), 3 ), sep = "-")
plot(pcm, type = "regions", names = lab)
## compare the cutpoints (with ylim specified manually)
opar <- par(no.readonly = TRUE)
ylim <- c(-2, 2)
layout(matrix(1:3, ncol = 1))
plot(pcm, type = "regions", parg = list(type = "mode"),
main = "Modus as Cutpoint", ylim = ylim)
plot(pcm, type = "regions", parg = list(type = "median"),
main = "Median as Cutpoint", ylim = ylim)
plot(pcm, type = "regions", parg = list(type = "mean"),
main = "Mean as Cutpoint", ylim = ylim)
par(opar)
## PCM for full verbal aggression data set
pcm_va <- pcmodel(VerbalAggression$resp)
plot(pcm_va, type = "regions")
if(requireNamespace("mirt")) {
## generalized partial credit model for full verbal aggression data set
gpcm_va <- gpcmodel(VerbalAggression$resp)
plot(gpcm_va, type = "regions")
}Please choose more modern alternatives, such as Google Chrome or Mozilla Firefox.