Please enable JavaScript.
Coggle requires JavaScript to display documents.
Psychometrics Module 11: Test bias, unfairness, and equivalence (test…
Psychometrics Module 11: Test bias, unfairness, and equivalence
-
establishing test bias
to test for possible slope bias, the interaction between age group and test performance (i.e. their cross product) would be entered in the second step of the regression equation. If the regression coefficient for the interaction term were significant, that would indicate a possible slope bias.
omitted variable problem: researchers wanted to see if omitting a variable from the analysis that was related to the criterion and the groupings variable, but not the other predictor (i.e. test scores) caused the predictor variable to appear to be biased against certain subgroups when, in fact, it was not.
test fairness
-
two individuals can take identical test, under identical circumstances and one person claim the test unfair and another person claim the test as fair.
unfairness comes into play when test takers are given advantages under the Americans with Disabilities Act
unlike test bias, it is not a psychometric concept
-
false negatives: individuals who were unsuccessful on the test but would have been successful on the criterion
-
constant ratio model: argued that there should be a constant ratio for each subgroup in terms of who is "successful" (on the criterion) and who passes the test
conditional probability: the proportion of individuals selected versus those who would be successful on the criterion should be the same for each subgroup
-
-
test equivalence
back translation:refers to tests that have been translated by bilingual individuals from English to the target language, and then retranslated back into English by other bilingual individuals
-
-
conceptual equivalence: examines whether, across cultures, the same meaning is attached to the terms used in the items
concerned with the degree to which test takers share a common understanding of the terms used in the items.
functional equivalence: determines the degree to which behavioral assessments function similarly across cultures.
scholar equivalence: assesses the degree to which different cultural groups produce similar means and standard deviation scores.
-
levels of accommodation
level 1: change in medium: present disabled individuals with the same test items presented to other test takers, but the items are presented in a different manner, such as providing a reader or a Braille version of a test to a blind individual
Level 2 accommodations: provide additional time to individuals with disabilities when completing a power test.
level 3 accommodations: "change in content" accommodations include item revision, deletion of items on a test, and change in item format