week 9 - bias Flashcards
test bias
arises when there is a factor inherent in a test that systematically prevents accurate, impartial measurement
- Bias in the test in regards to a particular group
construct bias
bias within the meaning of the test, two groups see the construct differently
eg, two groups of people see the construct of intelligence differently
predictive bias
bias in the predictive use of a test
eg, atar being baised towards western children as a predictor for academic success in uni
methods of detecting test bias
internal evaluations for construct bias
external evaluations for predictive bias
types of internal evaluations
detecting construct bias
Item Discrimination Index
Factor Analysis
Differential Item Functioning
Rank-Order Consistency
an item is biased if
- people belonging to different groups respond in different ways to the item, and
- if it could be shown that these differing responses were not related to group differences associated with the psychological construct measured by the test
item discrimination index
measures if the items correlate with the total score heavily
must use these seperately for both groups
is the item good and will it actually discriminate between the two groups
differential item functioning
an assesment of consruct bias which assumes it can measure trait levels directly from test data
in looking at the curve wed hope to see the same function between the two groups
rank order consistency
looking at the average rankings of responses to items
ensuring the order matches across groups
eg, item 3 is the highest scored item for men but second highest for woemn
rank order consistency
looking at the average rankings of responses to items
ensuring the order matches across groups
eg, item 3 is the highest scored item for men but second highest for woemn
external evaluations for test bias
looking at predicitive bais
bivariate regression
* slope bias
* intercept bias
* slope + intercept bias
* outcome score bias
2 steps in external evaluations
- you need to determine whether the test scores predict the dependent variable to start with
- you then need to determine whether the test scores predict the dependent variable equally well across groups
are they a good measure
are they a good measure across groups
outcome score bias
?