In SBGalvin/psketti: Generating Investigatory Plots and Tables for Rasch Analysis

knitr::opts_chunk$set(echo = TRUE)
library(knitr)

psketti

psketti is a package for generating investigatory plots and tables for Rasch Analysis. Models should first be estimated using eRm, psketti then pskettifies your Rasch data and produces outputs using ggplot2.

Installation

devtools::install_github("SBGalvin/psketti") # install

Loading

Loading psketti will produce the below startup message. If any of the dependencies are already loaded, psketti will tell you. If the dependencies are not installed at the time of the loading, psketti will tell you which ones to install.

library(eRm)
library(psketti)

-----------------------------------------------------
Loading psketti v 0.1.0
Beta Software !!!!


Dependencies:
    eRm      (1.0.2) already loaded
    dplyr        (1.0.4) already loaded
    ggplot2      (3.3.3) loaded by psketti
    viridis      (0.5.1) loaded by psketti


please install any NOT INSTALLED dependencies using:
install.packages('dependency')
-----------------------------------------------------

Investigate Item Response Functions - Dichotomous Rasch Model

1.1 Prepare data

data("FakeData") # load data
# restructure fake data
Fake_Data_scores <- reshape(FakeData[, c("ID", "Item", "X")],
                           timevar = "Item",
                           idvar = "ID",
                           direction = "wide")
# for eRm new names
names(Fake_Data_scores) <- c("ID",
                             paste0("i",
                                    sprintf(fmt  = "%02d", 1:23)))

row.names(Fake_Data_scores) <- Fake_Data_scores$ID
Fake_Data_scores$ID <- NULL

fake_rm   <- RM(Fake_Data_scores) # Estimate Rasch model

1.2 pskettify your data

Convert your data into a format for psketti to use. Select confidence levels for empirical class interval values for empirical Item Characteristic Curves (ICC), and set the minimum and maximum x axis values for the Rasch Item Response Function (IRF).

psk_data <-pskettify(eRm.obj = fake_rm, conf.level = .95, Theta.lwr = -6, Theta.upr = 6)

1.3 Plot Rasch IRF curves

Plot IRFs for individual items with empirical ICC, and class interval averages with confidence intervals

# plot ICC for one item
psk_1_present <- psketto(psk_data,
                         style = "present",
                         item = "i01",
                         item.label = "i01")
psk_1_present

psk_1_print <- psketto(psk_data,
                       style = "print",
                       item = "i01",
                       item.label = "i01")
psk_1_print

...or use psketti() to plot ICC for all plots. psketti() objects printed to the console will print suggestions for how to call a plot.

# plot ICC for one item
psk_IRF <- psketti(psk_data)
psk_IRF

psk_IRF$Plot.List[["i06"]][[1]]

psk_IRF$Plot.List[["i12"]][[1]]

1.4 Item Fit statistics

Produce a table of item fit statistics as per eRm

itemFit_psk <- item_fit_table(fake_rm)

# round ouput to 2 places
itemFit_psk[, -1] <- round(itemFit_psk[, -1], 2)
itemFit_psk

Produce Plot of infit and outfit statistics

psketti_msq(x = itemFit_psk)

1.5 Investigate distractor options

# response option categories
r_o <- factor(sort(unique(FakeData$K)), levels = sort(unique(FakeData$K)), ordered = TRUE)
tlt_data <- tabliatelle(x = FakeData, ID = "ID", Item = "Item", K = "K", 
                        response_options = r_o, eRm.obj = fake_rm)
tlt_data

1.6 Plot the distractor empirical ICC against the dichotomous Rasch ICC

# multiple plots
spag_plot <- psketti_distractor(ID = "ID",              # set ID column
                                Item = "Item",          # set Item column
                                K= "K",                 # Set resp categories 
                                x = FakeData,           # select data
                                eRm.obj = fake_rm,      # select eRm object
                                response_options = r_o, # set resp options
                                p.style = "present")    # set plotting style


spag_plot$Plot.List[['i01']][[1]] # plot item 1

1.7 Print a score report

K_opt <- factor(LETTERS[1:5], levels = LETTERS[1:5], ordered = TRUE)
score_report <- ingrediente(x = FakeData,
                            Item = "Item",
                            ID = "ID",
                            Score = "X",
                            K = "K",
                            K_options = K_opt,
                            Index = "Index")

# show score report for values with a total score <= 1
score_report[score_report$total_score <= 1, ]

2 Rasch Partial Credit Model

2.1 Prepare data

data("FakePCMData")

F2           <- FakePCMData
rownames(F2) <- F2$ID
F2$ID        <- NULL

fake_pcm <- PCM(F2)

2.2 pskettify

psk_pcm <- pskettify(eRm.obj = fake_pcm) # pskettify
psk_pcm 

2.3 Plot PCM ICC

psk_no_facet <- psketto(pskettified_data = psk_pcm, item = "i01", item.label = "i01")
psk_no_facet

2.3 Plot PCM ICC with facet curves

psk_facet <- psketto(pskettified_data = psk_pcm, item = "i01", item.label = "i01", facet_curves = TRUE)
psk_facet

2.4 plot all ICC

By default psketti() produces faceted plots for partial credit models

print_facet <- psketti(pskettified_data = psk_pcm)
print_facet$Plot.List[['i01']][[1]] # output

However, you can turn off this behaviour by using the Force_no_facet=True argument.

print_no_facet <- psketti(pskettified_data = psk_pcm, p.empICC = FALSE, Force_no_facet = TRUE)
print_no_facet$Plot.List[['i02']][[1]]

2.5 PCM item fit

Item fit statistic.

fake_ifit <- item_fit_table(eRm.obj = fake_pcm)

# round ouput to 2 places
fake_ifit[, -c(1, 2)] <- round(fake_ifit[, -c(1, 2)], 2)
fake_ifit

Infit and Outfit MSQ plot.

psketti_msq(x = fake_ifit)

SBGalvin/psketti documentation built on March 13, 2021, 1:47 p.m.