12-08-2014 02:39 AM
Journals which rely on APA or similar guidelines usually request eta-squared or partial eta-squared when reporting ANOVA. PROC MIXED does not have an option for reporting this, and the hand calculation may not be straightforward when covariance structures other than compound symmetry are used. Does anyone have an accepted general method for these calculations when using PROC MIXED? That is, other than re-doing the analysis in SPSS using "old style" methods to get an automatic calculation. (Submitting to another journal would be my preferred option, but I usually get outvoted.:smileysilly
The same style also calls for SDs rather than SEs for results summaries. For those journals which may be open to discussion, does anyone have a handy, authoritative reference which simply explains why least squares means and their associated SEs are more appropriate for interpreting ANOVA results?