5
$\begingroup$

I have two groups of ten observations each, where the weight are measured at three different points in time. Let´s say baseline, 2 months and 5 months after baseline.

Now I would like to draw a conclusion if the mean of one group is lower than the other group over time. They most likely weigh the same at the first measurement, with little variation.

There´s been a suggestion to fit a simple linear regression, where we use the difference in mean of the two groups as response and time as dependent. In my opinion it seems wrong, we will only have three observations. Would it be more correct to fit som kind of Repeated measurements ANOVA or any other suggestion? Thankful for any input.

$\endgroup$
2
  • 1
    $\begingroup$ Your's is a classic example for RM-ANOVA. You have 3 effects to check in it: (1) Between-groups difference combining the 3 times, (2) Within-subject difference, i.e. whether there is time effect combining the 2 groups, (3) Interaction between-within effect, i.e. whether 2 groups react differently through time. One side notion: if you deal with body weight, it is likely to be right skewed, so it's worth considering a transformation first, perhaps. $\endgroup$
    – ttnphns
    Commented Jul 3, 2012 at 6:56
  • $\begingroup$ Thank you for your response. I have an additional thought/question about the randomization. Only the obersvations/subjects has been randomly selected, could this be a problem when evaluating the model? for example time here cant be randomized and which group observations belongs isnt either randomized. Dont know if it makes sense, thankful for any response. $\endgroup$
    – Carlos
    Commented Aug 7, 2012 at 15:13

1 Answer 1

3
$\begingroup$

I work with clinical trial data a lot. A characteristic of such data is the repeating of measurements on subjects at a number of scheduled trial visits. We use repeated measures ANOVA to look for differences between groups of patients ( often comparing treatment groups with a control group). The regression appraoch doesn't make sense to me either. You have no interest in the slopes of the regression lines. Your only concern is how the means change over time.

$\endgroup$
2
  • $\begingroup$ Thanks for your reply. I think that the idea was that we know one groups mean will increase over time, and the other group wont increase or dont increase as much. The idea is to show that the slope of one group differ from the other. Anyways, I will look into repeated measurements more thoroughly. Regards $\endgroup$
    – Carlos
    Commented Jul 2, 2012 at 14:56
  • $\begingroup$ Is it okay if some samples don't have multiple time points? For example, A - 1point, B,C - 2 points. $\endgroup$
    – Crispy13
    Commented Aug 3, 2023 at 5:11

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.