1
$\begingroup$

I am currently working on a small observational and retrospective study with 2 groups with different types of treatment. We have 3 outcome measures for both groups. The outcome measures were obtained at different times after the treatment occurred. I am not sure how to correct for the fact that measures were obtained at different times after treatment occurred. Would it be possible to do a linear regression with time as a predictor to ascertain that time does not have a significant impact on the outcome measure that i'm interested in? Would this justify comparing means between the two groups with an unpaired t-test?

Improve this question
$\endgroup$

1 Answer 1

Reset to default
0
$\begingroup$

You could create a dummy variable where people in group 1 get a zero and people in group 2 get a 1. You could do something more fine grained as well, say, maybe the minutes between tests if you think that is important.

By observational study, do you mean that this is not an "experiment" (e.g., random assignment to treatment and random selection)? If not an experiment, remember that you could find this new "time" variable as being significant (or not), but those results could change if you controlled for every other confounder which is omitted. I believe t-tests have independence assumptions and that the difference between the groups is distributed normally, but do not quote me on that.

Improve this answer
$\endgroup$
2
  • $\begingroup$ By observational I mean that there we do not do the treatment selection, we merely look at functional outcomes of treatments that were assigned by treating physicians and not by a research group based on a random selection or protocol otherwise $\endgroup$
    – Rachelle Bent
    Commented Feb 4, 2021 at 16:07
  • $\begingroup$ Ok, so in other words it seems like non-random assignment to treatment and control and non-random selection? In that case, you could fit a linear model and check the outcome of the dummy variable time but you have to remember that unobserved confounders remain uncontrolled so you probably should not bet much on its results. $\endgroup$
    – John Stud
    Commented Feb 4, 2021 at 16:58

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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