JMetrik User Guide mentions that "Item-total correlation type – Select Pearson correlation to have jMetrik compute Pearson correlations (point-biserial for binary items, Pearson correlation for polytomous items)." Can we interpret the output of polytomous items in the similar manner as the output of point-biserial correlation coefficient of dichotomous items? I need references to support the interpretation of discriminating power index for polytomous items. Thank you.
Hi
In classical test theory(C.T.T), I do not know the Discriminant index for polytomous items, But I suggest you do your analysis in the Item-Response theory(IRT). polytomous item response models are able to define and easily interpret polytomous items. follow the below link:
https://www.ets.org/research/policy_research_reports/publications/report/1996/ibtw
Hello Ibnu,
The point-biserial correlation referred to is a simple way to express the strength of association between scores on a dichotomous item with a (usually, corrected) total score. This is referred to as a discrimination index for the item.
In point of fact, the PB correlation is just a simplified, but numerically equivalent, version of the ordinary Pearson product-moment correlation, that was developed mainly to speed up hand calculation of values (before the age of computer software to do the heavy lifting).
As a correlation is just a correlation, you may certainly do likewise with polytomously scored items (e.g., compute a Pearson correlation). If you are at all concerned about whether the intervals are equal for either item scores or total scores, you could use either a Spearman correlation or make an assumption and use a polychoric correlation.
Good luck with your work.
- Badges
- Science topic
- Similar topics
- Mathematics