Data Format for Repeated Measures AnoVa/SPAnoVa

Hi everyone, I am a newcomer to the forum, SPSS and stats in general. Having followed the inbuilt tutorials in SPSS (V15) and using Julie Pallant's excellent Survival Guide, I am having problems arranging my data for analysis, especially with what SPSS expects for each fields in the Analyse -> GLM -> Repeated Measures creation wizard.

For the data collection phase of my project, I have data from 21 subjects, each with:

  • 3 measures of a parameter prior to applying a treatment
  • 3 measures of the same parameter following the treatment

What I am trying to find is:

  1. Do the "before" Vs. "after" treatment measures vary significantly (i.e. does the treatment alter the parameter of interest significantly, within-subjects)
  2. Do the "before" and, separately, "after" measures show significant variance across all subjects (i.e. do all subjects show similar "before" and "after", between-subjects values)

Showing my 1st 2 subjects' data only, my formatting is:
|____1____|__Absent___ |____40.80___ |___40.30___|___39.90____|
|____1____|__Present___|____42.10___ |___42.00___|___42.30____|
|____2____|__Absent___ |___41.20____ |___43.10___|___42.90____|
|____2____|__Present___|___41.30____ |___43.00___|___42.76____|

I am unsure if I have arranged suitably for the test I am using, what SPSS classifies as Within-Subject Factor Name(s), definitions and the number of levels for this as well as what to specify (if anything) for Measure Name. Any advice on whether the test I am considering is the optimum for what I am trying to discover from my data.

Any help/advice would be appreciatively-received.


Very nicely presented & easy to follow document, thanks.

At first glance, I thought that the model described was precisely what I was looking for, however, upon closer inspection (and abstracting the case study to my study) there are subtle differences which makes me think that the RM AnoVa may not be the test I am looking for.

The guide describes how the results are organised: The Within-Subjects factor is weight before, during and after a diet and the Between-Subjects factor is gender, namely Male and Female.

The test looks at 3 things:
  1. The effect of the diet in time (Within-Subjects factor)
  2. A Gender effect (Between-Subjects factor )
  3. The interaction of the within/between factors - i.e. does gender affect weight loss in time.

Now, in my study, the Within-Subjects factor are the trials, the before/after treatment is the (Between-Subjects factor. Applying this to the model given, the studies at first seem directly comparable:

But, there are a few issues that I can see:

Whereas, in the diet study the time effect of the diet was of interest, in my study, the 3 trials were taken to give a more robust idea of the parameter; the effect in time is not an issue itself. The guide study seems to suggest that the repeated measures are primarily concerned with measuring an effect in time - namely weight.

More importantly, the Between-Subjects factor in the guide is an actual separator of subjects, whereas in my study, the treatment is separating data which comes from the same subject in each case.

Thinking about it in basic terms, what I want to do is:

For each subject:
Evaluate if the "before treatment" data varies significantly.
Evaluate if the "after treatment" data varies significantly.
Evaluate if the before and after data values vary significantly.

then, repeat the same, but evaluate the variances across the entire population. i.e. variance within trials, between before/after treatment and then across the entire population.


TS Contributor
you cannot assume the before and after between subjects factor. They are the same people. I do not have spss in front of me, so i will say do this: subract from after the before and simply perform R-M anova with your within subjects factor only.