Skip to contents

mirtverse sits on top of mirt and leaves estimation alone. What it adds is the work around a model: Tidy data frames for your pipeline, ggplot2 figures you can restyle, and reports you can attach to a paper.

Your path in three steps

1. Keep using mirt, get tidy outputs. Any mirt() model works:

library(mirtverse)
mod <- mirt(data, 1, itemtype = "2PL", verbose = FALSE)
tidy(mod)       # one row per parameter, with standard errors
glance(mod)     # one row of fit statistics
augment(mod)    # person scores added to the data

2. Make figures for the paper. Each gg_*() function returns a ggplot2 object, and theme_mirtverse() gives a clean journal style:

gg_trace(mod, which.items = 1:4)
gg_wright(mod)
gg_q3(mod)          # local dependence

See the Gallery for every figure type with its code.

3. Go beyond the standard model. Response-time models, DIF in one call, and explanatory item models from item text:

fit_rt(mod, rt)                      # joint responses and response times
dif(data, group = group)             # DIF with effect sizes
d <- merge(itemtext, item_features(itemtext))
predict_difficulty(d, c("words", "max_digits", "decimals"))  # difficulty from item text

Citing

Please cite both mirt and mirtverse. citation("mirtverse") returns both references; the authors page has them too, and the book can be cited separately (10.5281/zenodo.23885605).