IRT 1 Flashcards
graded response model
ordered categorical responses
b is a category parameter instead of an item parameter
probability of answering in a certain category or higher
partial credit model
ordered categorical responses
b is at the intersection between adjacent categories
probability of answering in a certain category or higher, but subjects can obtain 0, 1, 2, or 3 points for a given question (high on LV means you will score high on the correct answer)
nominal response model
unordered categories
b and a are category parameters instead of item parameters
probability of answering in a certain category or higher, but subjects can obtain 0, 1, 2, or 3 points for a given question (high on LV means you will score high on the correct answer)
item information
amount of information that an item provides for a given theta.p
the item with the largest information is the item with the largest discrimination
the maximum information of this item is at it’s difficulty value
the information at this point is 1/4 * a^2
maximum information for 1 and 2 parameter models
maximum information is at the item’s difficulty parameter
1/4 * a.i^2
maximum information for the 3 parameter model
maximum information is at theta.max
test information
amount of information that a test provides for a given theta
sum of all item information values
SE(theta-hat)
1/sqrt(test information)
target information
desired information, specified by the researcher
range of scores that are included in the information obtained
information for multiple category models (GRM, PCM, NRM)
each response category gives information
degree of parameter bias
reflected by the amount by which the estimates depart systematically from the identity line
degree of parameter imprecision
standard error
reflected by the amount by which the estimates scatter around the identity line
GRM contra-indicative items
show a plot with categories in reverse order
contribution of item information to test information
to see if an item contributes a significant amount of information to the test infromation, you can check whether the discrimination parameter is significant for this item.
