I'm new here but I thought y'all would be the best place to ask.

Is there some research sub-community that studies how to measure psychological constructs and processes when you have high-resolution behavioral data? I'm vaguely aware of psychometrics but I've only heard of it in the context of designing questionnaires.

To avoid the XY problem, let me describe what I mean a little more. I think I've got a handle on psychometrics as [broadly] the science of measuring psychological states and constructs. In contrast to the questionnaires I often see, I have behavioral data - in particular, I have virtual reality position tracking data. Participants in our experiments wear the headset and use hand controllers and we get position and rotation data at 90Hz over the course of the 5-20 minute experiment. I say this is "high-resolution behavioral data" because it is behaivoral (rather than self-report-based) and high-resolution because the data collected is not a single measure (e.g, how close did someone approach an object in VR?)

There are a lot of ways to measure any proposed construct. For example, one of our rough observations over five-ten years of runnign studies is that some people get super into VR and move around a lot while others don't move around as much. Some explore scenes and others don't. There are a bunch of ways that have been used in the past (standard deviation of yaw angle, total amount of movement, number and average length of fixations, amount of horizontal range that has been 'seen', amount of sphere being seen, etc. [apologies if those don't make sense as written, the key idea here is that they all correlate at about 0.5-0.8 but they legitimately measure different things.

I'd like to be able to evaluate them coherently, and of course that means using previously invented frameworks if possible. I'd like to ask questions like: how reliable are they? how related are they? do they measure different things? are they measuring different parts of the same thing? which are best related to X construct of interest? which are related only because of noise? how do you account for noise and multiple comparisons when your measures are highly correlated?

I'd also like to answer other questions similar to those that I haven't thought of - and I assume the psychometrics community (or some sub-field of it) would have converged on the good questions.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.