anchor function provides a variety of anchor
methods for the detection of uniform differential item functioning (DIF)
in the Rasch model between two pre-specified groups. These methods can
be divided in an anchor class that determines characteristics of the
anchor method and an anchor selection that determines the ranking order
of candidate anchor items. The aim of the anchor function is to
provide anchor items for DIF testing, e.g. with
anchortest.
anchor(object, ...)
"anchor"(object, object2, class = c("constant", "forward"), select = NULL, length = NULL, range = c(0.1, 0.8), ...)
"anchor"(formula, data = NULL, subset = NULL, na.action = NULL, weights = NULL, model = raschmodel, ...)raschmodel.anchor.default in the formula method. In the
default method, these additional arguments are currently not being used.constant anchor class implying a constant anchor length
defined by length and the iterative forward anchor
class, for an overview see Kopf et al. (2014a)."MTT", "MPT", "MT", "MP",
"NST", "AO", "AOP". For details see below and
Kopf et al. (2014b). Following their recommendations, the default
selection = "MPT" is used for class = "constant" whereas
selection = "MTT" is used for class = "forward".forward anchor grows up to the proportion
of currently presumed DIF-free items specified in range
and the constant anchor class selects four anchor items,
unless an explicit limiting number is defined in length
by the user.forward anchor class is used and the
second element determines a percentage of currently presumed DIF-free
items up to which the anchor from the forward anchor class is
allowed to grow.y ~ x where y specifies a
matrix of dichotomous item responses and x the grouping
variable, e.g., gender, for which DIF should be tested for.formula.NAs).itempar
method, by default raschmodel.anchor, i.e. a list including
In the constant anchor class, the anchor length is pre-defined by the
user within the argument length. The default is four anchor
items. The iterative forward class starts with a single anchor item and
includes items in the anchor as long as the anchor length is shorter
than a certain percentage of the number of items that do not display
statistically significant DIF (default: 0.8). Furthermore, a percentage
of first anchor candidates is excluded from consideration (default: 0.1)
and the user is allowed to set a maximum number of anchor items using the
argument length. A detailed description of the anchor classes can be found
in Kopf et al. (2014a).
Both anchor classes require an explicit anchor selection strategy (as opposed to
the all-other anchor class which is therefore not included in the
function anchor). The anchor selection strategy determines the ranking order of
candidate anchor items. In case of two groups, each item $j, j = 1, \dots, k$ (where $k$ denotes the number of
items in the test) obtains a criterion value $c_j$ that is
defined by the anchor selection strategy. The ranking order is
determined by the rank of the criterion value
rank$(c_j)$.
The criterion values are defined by the respective anchor selection strategy as in the following equations, where $1$ denotes the indicator function, $t_j$ denotes the test statistic for item $j$, followed by the underlying anchor items in parenthesis, $p_j$ the corresponding p-values (also with anchor items in parenthesis), $<0.5k>$ denotes the empirical 50% quantile and $A_{purified}$ the anchor after purification steps (a detailed description of the anchor selection strategies implemented in this function is provided in Kopf et al., 2014b).
All-other selection by Woods(2009), here abbreviated AO: $$ c_j^{AO} = | t_j (\{1,\ldots,k\}\textbackslash j) |$$
All-other purified selection by Wang et al. (2012), here abbreviated AOP: $$ c_j^{AOP} = | t_j ( A_{purified} ) |$$
Number of significant threshold selection based on Wang et al. (2004), here abbreviated NST: $$ c_j^{NST} = \sum_{l \in \{1,\ldots,k\} \textbackslash j} 1 \left\{ p_j ( \{l\} ) \leq \alpha \right\}) |$$
Mean test statistic selection by Shih et al. (2009), here abbreviated MT: $$ c_j^{MT} = \frac{1}{k-1} \sum_{l \in \{1,\ldots,k\} \textbackslash j} \left| t_j ( \{l\}) \right| $$
Mean p-value selection by Kopf et al. (2014b), here abbreviated MP: $$ c_j^{MP} = - \frac{1}{k-1} \sum_{l \in \{1,\ldots,k\} \textbackslash j} p_j ( \{l\} ) $$
Mean test statistic threshold selection by Kopf et al. (2014b), here abbreviated MTT: $$ c_j^{MTT} = \sum_{l \in \{1,\ldots,k\} \textbackslash j} 1 \left\{ \left| t_j ( \{l\} ) \right| > \left( \left| \frac{1}{k-1} \sum_{l \in \{ 1, \ldots, k \} \textbackslash j} t_j ( \{l\} ) \right| \right)_{\left( \lceil 0.5\cdot k\rceil \right)} \right\} $$
Mean p-value threshold selection by Kopf et al. (2014b), here abbreviated MPT: $$ c_j^{MPT} = - \sum_{l \in \{1,\ldots,k\} \textbackslash j} 1 \left\{ p_j ( \{l\} ) > \left( \frac{1}{k-1} \sum_{l \in \{ 1, \ldots, k \} \textbackslash j} p_j ( \{l\} ) \right)_{ \left( \lceil 0.5\cdot k\rceil \right)} \right\} $$
Kopf et al. (2014b) recommend to combine the class = "constant" with
select = "MPT" and the class = "forward" with select = "MTT",
respectively.
The all-other anchor class (that assumes that DIF is balanced i.e. no
group has an advantage in the test) is here
not considered as explicit anchor selection and, thus, not included
in the anchor function (but in the anchortest
function). Note that the all-other anchor class requires strong prior
knowledge that DIF is balanced.
Julia Kopf, Achim Zeileis, Carolin Strobl (2014b). Anchor Selection Strategies for DIF Analysis: Review, Assessment, and New Approaches. Educational and Psychological Measurement, accepted for publication.
Ching-Lin Shih, Wen-Chung Wang (2009). Differential Item Functioning Detection Using the Multiple Indicators, Multiple Causes Method with a Pure Short Anchor. Applied Psychological Measurement, 33(3), 184--199.
Wen-Chung Wang (2004). Effects of Anchor Item Methods on the Detection of Differential Item Functioning within the Family of Rasch Models, Journal of Experimental Education, 72(3), 221--261.
Wen-Chung Wang, Ching-Lin Shih, Guo-Wei Sun (2012). The DIF-Free-then-DIF Strategy for the Assessment of Differential Item Functioning. Educational and Psychological Measurement, 72(4), 687--708.
Carol Woods (2009). Empirical Selection of Anchors for Tests of Differential Item Functioning. Applied Psychological Measurement, 33(1), 42-57.
anchortest
## Verbal aggression data
data("VerbalAggression", package = "psychotools")
## Rasch model for the self-to-blame situations; gender DIF test
raschmodels <- with(VerbalAggression, lapply(levels(gender), function(i)
raschmodel(resp2[gender == i, 1:12])))
## four anchor items from constant anchor class using MPT-selection
canchor <- anchor(object = raschmodels[[1]], object2 = raschmodels[[2]],
class = "constant", select = "MPT", length = 4)
canchor
summary(canchor)
## iterative forward anchor class using MTT-selection
set.seed(1)
fanchor <- anchor(object = raschmodels[[1]], object2 = raschmodels[[2]],
class = "forward", select = "MTT", range = c(0.05, 1))
fanchor
## the same using the formula interface
set.seed(1)
fanchor2 <- anchor(resp2[, 1:12] ~ gender , data = VerbalAggression,
class = "forward", select = "MTT", range = c(0.05, 1))
## really the same?
all.equal(fanchor, fanchor2, check.attributes = FALSE)
Run the code above in your browser using DataLab