Predictive validity

Pretty much what it says: do measurements on some measure predict future events as theory about the variable measured would suggest. I’m not sure I’ve ever seen a formal test of predictive validity of an therapy/MH/WB change measure which is pretty sad really when there are surely thousands of measures that are claimed to be “validated measures” and tens of thousands of papers claiming to have validated a measure by reporting a factor analysis of its item scores in some dataset.

Details #

I’m not saying that factor analyses of measures are not somewhat informative about how the items of a measure performed in that dataset, just that for some measures that’s not an attribute that the designers, or the users, of the measure really needed. (Yes, I’ve got skin in this game as that’s very true for the CORE measures, but it’s true for many others certainly including the OQ measures.)

Predictive validity is a very strong test of a measure and it’s a bit resource intensive as, as the name suggests, it needs people to be followed or recontacted over time. It’s also not one of those neat things where we have an expectation of very high predictive strength: what future consequences of scores on most of our measures would we expect to be very strongly predicted by earlier scores? However, in the developmental psychology realm I believe there are measures, for example of attachment status and probably other postulated variables that have been shown to have some predictive correlations with later events.

It would be lovely to see much more exploration of predictive validity for adult change measures: do termination scores at the end of therapies predict relapse rates?

Try also #

Convergent validity
Divergent/discriminant validity

Chapters #

Touched on in Chapters 3 and 7.

Online resources #

This largely boils down to test for correlations so various of my shiny apps relating to correlation are pertinent I guess.

Dates #

First created 22.iv.24.

Powered by BetterDocs