MATLAB Answers

Confusing documentation for analysis of variance (ANOVA)

7 views (last 30 days)
z8080
z8080 on 10 Mar 2017
Commented: z8080 on 1 Sep 2017
Reading the documentation for Matlab's repeated-measures ANOVA, I noticed two things that didn't make sense to me (I am reasonably literate in statistics):
1) In this example (the fisheriris dataset), "species" is taken as the predictor variable - however species is a between-subjects factor, not within-subjects!
2) The output of another example for a repeated measures ANOVA lists Group as a within-subjects factor when it too is clearly between-groups.
Also, why is it that all factors' names in that output table end in "time", e.g. "Group:Time"? Is that because most repeated measures are normally measurements done across time? This is really confusing.
Thanks for any help clarifying these points!

  0 Comments

Sign in to comment.

Answers (1)

David Resendes
David Resendes on 21 Apr 2017
RE (1): In this example , observations are irises. The measurement that can vary among different irises is its species, so it should be a predictor variable treated as a between-subjects factor. You seem to be saying that the example treats it as a within-subjects factor. Can you point to the place in the example where species is treated as a within-subjects factor?
RE (2): The link you provided doesn’t take me to another example. However, you might be referring to this example . In this example, the second sentence states that Group is a between-subjects factor, and it is treated as a between-subjects factor in the ANOVA table. Can you point me to the place in this example, or to another example, where a variable called Group is treated as a within-subjects factor?
RE your final question: The description of the first output argument of ranova states:
ranovatbl includes a term representing all differences across the within-subjects factors. This term has either the name of the within-subjects factor if specified while fitting the model, or the name Time if the name of the within-subjects factor is not specified while fitting the model or there are more than one within-subjects factors. ranovatbl also includes all interactions between the terms in the within-subject model and all between-subject model terms.
In this example, there are two within-subject factors, so Time is used as the variable’s name.

  1 Comment

z8080
z8080 on 1 Sep 2017
Sorry for the very late reply!
1) The example does treat is a between-subjects factor, but why I found that surprising is because the model itself that is fitted is not a between-subjects (or mixed) model, but a repeated measures (i.e., within-subjects) model.
Unfortunately I cannot access your links, as the website asks me for a licence number ("You are not associated with a license or a trial.To view the documentation for this product, associate to a license or contact sales. "), which I don't have, despite Matlab starting fine on my work PC (which I guess means our group's license is still valid).

Sign in to comment.

Sign in to answer this question.