1
$\begingroup$

Say I have a go/no-go task and my output data includes 3 parameters: average reaction time, variance of reaction time, and number of errors. I want to composite all of the parameters into a single index score.

What is the best way to do it? All of the parameters are equally important to me.

$\endgroup$
  • $\begingroup$ Welcome to Psychology.SE! I am not sure by reading this question that it has anything to do with Psychology or Neuroscience. This seems to be a stats question to me which would suit CrossValidated.SE more than here. Is there more you can provide to show it is on topic here? $\endgroup$ – Chris Rogers Jul 25 '18 at 21:36
  • $\begingroup$ @ChrisRogers go/no go and reaction time measures are predominantly used in psychology. So, I think it is in scope. $\endgroup$ – Jeromy Anglim Jul 26 '18 at 5:37
  • 1
    $\begingroup$ @ChrisRogers as said over in meta, this is ontopic. $\endgroup$ – AliceD Jul 26 '18 at 7:34
  • 1
    $\begingroup$ I think the difference is that how to create this composite is grounded in psychological theory. I.e., it's not just a generic question about how to do a t-test. It's about how to score a psychological measure, where the rule for scoring depends on the psychological meaning of the components. Conference presentations and journal articles in mathematical and cognitive psychology are devoted to the topic of how to trade-off speed and accuracy in reaction times. $\endgroup$ – Jeromy Anglim Jul 26 '18 at 8:20
1
$\begingroup$

This overlaps a lot with these two questions: So check them out

Regarding some specifics of your question, reaction time variance will often be strongly correlated with reaction time mean. So in routine measures of individual differences where speed is the main focus, it may be sufficient to focus on mean reaction time and accuracy. E.g., mean reaction time possibly with some sort of penalty for errors (see the above answers).

If you are wanting to get very precise, then there are a range of modelling approaches that may be relevant (e.g., this question on LBA and diffusion models).

With regards to go-no-go, I believe there are fairly standardised procedures for scoring such tests.

$\endgroup$
  • 1
    $\begingroup$ If you want to model go-nogo with the LBA there's tutorials for doing that here: osf.io/pbwx8 $\endgroup$ – steveLangsford Jul 26 '18 at 15:03

Your Answer

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

Not the answer you're looking for? Browse other questions tagged or ask your own question.