difstudy: Differential Item Functioning Study

In talbano/epmr: Tools for Educational and Psychological Measurement

Differential Item Functioning Study

Description

Functions for estimating differential item functioning (DIF) with a set of scored item responses.

Usage

difstudy(
  x,
  groups,
  ref,
  scores = NULL,
  method = c("mh", "lr"),
  anchor_items = 1:ncol(x),
  dif_items = 1:ncol(x),
  complete = TRUE,
  na.rm = FALSE,
  p_cut = 0.05,
  sdp_w = c("focal", "reference", "total"),
  contrast_type = "contr.treatment"
)

Arguments

x	matrix or data.frame of scored item responses.
groups	vector defining categorical grouping variable.
ref	string identifying label used in groups to represent the reference group, with no default.
scores	optional vector of construct scores, defaulting to row sums over all columns in x when NULL and anchor_items is not specified.
method	what DIF method to use, with "mh" for Mantel-Haenszel and "lr" for logistic regression.
anchor_items	vector of item names or column numbers for identifying anchoring items in x. These will be used when finding construct scores if scores is missing, defaulting to all columns in x.
dif_items	vector of item names or column numbers identifying what items in x should be analyzed for DIF, defaulting to all columns in x.
complete	logical with default TRUE indicating whether or not x, groups, and scores should be reduced to cases with complete data.
na.rm	logical with default FALSE specifying whether missings should be removed before calculating individual descriptives.
p_cut	numeric cutoff for evaluating the p-value for the Mantel-Haenszel chi-square, with default 0.05.
sdp_w	string for choosing the weight when finding the standardized difference statistic, defaulting to the number of test takers in the focal group.
contrast_type	string for the type of contrasts to use in logistic regression, defaulting to "contr.treatment".

Examples

# Vector of item names
ritems <- c("r414q02", "r414q11", "r414q06", "r414q09",
  "r452q03", "r452q04", "r452q06", "r452q07", "r458q01",
  "r458q07", "r458q04")
rsitems <- paste0(ritems, "s")
# Subset PISA09 to Belgium and Germany
pisa <- subset(PISA09, cnt %in% c("BEL", "DEU"))
x_totals <- rowSums(pisa[, rsitems], na.rm = TRUE)
# Uniform DIF with Mantel-Haenszel
difstudy(x = pisa[, rsitems], groups = pisa$cnt, ref = "BEL",
  scores = x_totals, method = "mh")
# Uniform and nonuniform DIF with logistic regression
difstudy(x = pisa[, rsitems], groups = pisa$cnt, ref = "BEL",
  method = "lr")

talbano/epmr documentation built on May 1, 2024, 11:10 a.m.