Dear all,
I am trying to compare the predictive utility of two measures. Both measures
are intended to tap the same unidimensional construct (Variable X), which is
expected to predict Outcome Z, when controlling for Variables Y1, Y2 etc.
Both measures are made up of multiple items, and the score on each measure
represents the mean of the items. The score for one of the measures is the
mean of all Variable X items, whereas the score on the other measure
represents the mean of a subset of those items. (So, one measure is a
subscale of the other). The data I have are from a single sample which has
completed all items (i.e. both measures) (and Variables Y1, Y2 etc). The two
measures are (unsurprisingly) highly correlated (.90+). I don’t have the raw
item scores, only the means of each measure. Is there any way in which I can
test statistically the hypothesis that one measure is better than the other at
predicting Outcome Z?
Best wishes,
Ben Gardner
|