GDINA (version 2.0.8)

itemfit: Item fit statistics

Description

Calculate item fit statistics (Chen, de la Torre, & Zhang, 2013)

Usage

itemfit(GDINA.obj, person.sim = "post", p.adjust.methods = "bonferroni",
  digits = 4, N.resampling = NULL, randomseed = 123456)
# S3 method for itemfit
extract(object, what, ...)
# S3 method for itemfit
heatplot(object, ...)
# S3 method for itemfit
summary(object, ...)

Arguments

GDINA.obj

An estimated model object of class GDINA

person.sim

Simulate expected responses from the posterior or based on EAP, MAP and MLE estimates.

p.adjust.methods

p-values for the proportion correct, transformed correlation, and log-odds ratio can be adjusted for multiple comparisons at test and item level. This is conducted using p.adjust function in stats, and therefore all adjustment methods supported by p.adjust can be used, including "holm", "hochberg", "hommel", "bonferroni", "BH" and "BY". See p.adjust for more details. "bonferroni" is the default.

digits

How many decimal places in each number? The default is 4.

N.resampling

the sample size of resampling. By default, it is maximum of 1e+5 or ten times of current sample size.

randomseed

random seed; This is used to make sure the results are replicable. The default random seed is 123456.

object

objects of class itemfit for various S3 methods

what

argument for S3 method extract indicating what to extract; It can be "p" for proportion correct statistics, "r" for transformed correlations, logOR for log odds ratios and "maxitemfit" for maximum statistics for each item.

...

additional arguments

Value

an object of class itemfit consisting of several elements that can be extracted using method extract. Components that can be extracted include:

p

the proportion correct statistics, adjusted and unadjusted p values for each item

r

the transformed correlations, adjusted and unadjusted p values for each item pair

logOR

the log odds ratios, adjusted and unadjusted p values for each item pair

maxitemfit

the maximum proportion correct, transformed correlation, and log-odds ratio for each item with associated item-level adjusted p-values

Methods (by generic)

• extract: extract various elements from itemfit objects

• heatplot: plot bivariate heatmap for misfit detection

• summary: print summary information

References

Chen, J., de la Torre, J., & Zhang, Z. (2013). Relative and Absolute Fit Evaluation in Cognitive Diagnosis Modeling. Journal of Educational Measurement, 50, 123-140.

Examples

# NOT RUN {
dat <- sim10GDINA$simdat
Q <- sim10GDINA$simQ

mod1 <- GDINA(dat = dat, Q = Q, model = "GDINA")
mod1
itmfit <- itemfit(mod1)

# Print "test-level" item fit statistics
# p-values are adjusted for multiple comparisons
# for proportion correct, there are J comparisons
# for log odds ratio and transformed correlation,
# there are J*(J-1)/2 comparisons

itmfit

# The following gives maximum item fit statistics for
# each item with item level p-value adjustment
# For each item, there are J-1 comparisons for each of
# log odds ratio and transformed correlation
summary(itmfit)

# use extract to extract various components
extract(itmfit,"r")

mod2 <- GDINA(dat,Q,model="DINA")
itmfit2 <- itemfit(mod2)
#misfit heatmap
heatplot(itmfit2)
itmfit2
# }