1

This is a question that has been bothering me for a long time, and until now I haven't been able to come up with a decent solution.

Imagine you do a straightforward repeated measures analysis in SPSS with your data in the wide format (necessary for GLM repeated measures). In this analysis, if you request descriptives, SPSS provides you with the descriptives of the conditions (i.e., the columns in the wide format). So if you have a 2 by 2 repeated measures anova, you get four condition means and SDs.

However, I am also interested in the means and SDs per independent variable, so collapsed over the levels of the other variable. Of course SPSS provides estimated marginal means, but these are not the "true" descriptives (they are, well, estimates, and provide SE instead of SD).

Is there a way, in the GLM repeated procedure or otherwise to get this information?

Many thanks, Martijn

Martijn Goudbeek
  • 11
  • 1
  • The marginal mean is the sample mean. To get the sample standard deviation, why not just multiply the standard error by the square root of the sample size for that group? – bsbk Dec 03 '14 at 15:28
  • Because SPSS does not present you with the "marginal mean", but with the "Estimated Marginal Mean", which is (in SPSS GLM) the mean response for each factor, _adjusted for any other variables in the model_. This is, to me, not identical to the "raw" descriptive statistics, which is what I'm after. – Martijn Goudbeek Dec 04 '14 at 11:00
  • See also [here](http://stats.stackexchange.com/questions/41789/what-makes-a-glm-estimate-the-means-differently-from-the-actual-sample-means) – Martijn Goudbeek Dec 04 '14 at 11:07
  • I see. Well, this is not an elegant solution but one way would be to convert from wide format to long format, split the file based on a main effect, and then run descriptives. Then repeat with the file split by the other main effect. – bsbk Dec 04 '14 at 15:29
  • That is an interesting suggestion. However, would the SD be calculated in the same way as in a repeated measures design? I don't think it would, actually. – Martijn Goudbeek Dec 07 '14 at 10:00

0 Answers0