Abstract
We tested if a propensity score (PS) matching method supports the unidimensionality assumption of the Rasch model which is critical to link similar rehabilitation instruments. We obtained 1,013 respondents from the 2009 Hispanic Established Populations for Epidemiologic Studies of the Elderly Frailty study. We used a unidimensional item pool of 10 SF-36 physical function and nine activities of daily living items. Subjects were matched based on their functionality (high and low), and exploratory factor analysis was used to test if the item pool in the matched sample holds the unidimensionality assumption. The study findings revealed that the matched sample demonstrated two distinct measurement structures with excellent model fit. This finding indicates that the PS matching did not mimic the common-person assumption. Therefore, the combination of PS matching and common-person equating method may not be appropriate to equate two rehabilitation-related instruments administered to two different groups.