I know the Repeated measures ANOVA is the equivalent of the one-way ANOVA, but for related, not independent groups but I can't understand. When can I use this design? When one way? And when repeated design?
$\begingroup$
$\endgroup$
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
4
Nomen est omen: You have to use repeated measures procedures, if you have multiple observations on the same experimental unit (e.g. proband), just at different time points or places. This happens usually, if there is at least one (fixed) factor like time crossed with the other factors. It is recommended to study the terminology of factorial designs first. Then it is easy to see if an experiment requires repeated measures analysis.
If you don't want to infer anything about the time factor, you can calculate the mean of the time points per subject and do a one-way ANOVA on these means.
-
$\begingroup$ Ok It's right but I guess that We can use anova or block design When we have different subject per any time or condition. It means we need more subject for design.For example 4 subject but by same features for 4(level) time or condition point not one subject for all 4 level of time or condition. $\endgroup$ Commented Jan 9, 2015 at 9:58
-
$\begingroup$ Yes, in this case you don't have repeated measures. In terms of factorial designs, your example means that the subjects are nested under the time or condition, that means each subject has an observation only at a certain time point. If on the other hand subject and time are crossed, each combination of subject and time exists. These combinations are the repetitions. $\endgroup$ Commented Jan 9, 2015 at 11:05
-
$\begingroup$ The key question for me is still regarding "If you don't want to infer anything about the time factor [...]" If you do want to infer something about the time factor, wouldn't this imply a possibly totally different design and/or test optimized to test time specifically? Does just because you have multiple measurements justify a repeated measures ANOVA? In my case I take 8 measurements over the course of 50 minutes, however, I do not control fully for the nature of them (they differ slightly). I take 8 measurements exactly because I want to have a meaningful mean. What if I would take 100? $\endgroup$ Commented Jan 22, 2015 at 18:18
-
$\begingroup$ @StevenJeuris... the design of the model is generally restricted by how you collect your data and the type of data it is. As such, if the dependent variable is measured at different time points on the same experimental unit, then the model will be restricted to some form of repeated measures design. In your case, if the 8 measurements are not taken at the same time for each person, a repeated measures ANOVA cannot account for the differences in timing (but still might work well), but a mixed model with time as a random effect might be better suited. Still a repeated measures mixed model $\endgroup$– MensenCommented Jan 22, 2015 at 20:10