I have conducted an analysis of covariance, ANCOVA, followed by Bonferroni post-hoc tests, on SPSS. I want to report the differences between my groups in text (APA style) and in a corresponding bar graph. My covariate (a different variable from the DV and IV, not a pretest) was found significant, so the descriptive means and the estimated marginal means (EMM) are different.

I want to know which of the means ("descriptive" or EMM) is better to report and put in the graph. One way to think about it is that the adjusted means are not real data but simply estimations from the software --nothing my article should be about. However my analysis will be drawn on the conclusions of the ANCOVA, comparing the EMM, so I should focus on the EMM.

I have explored various manuals and articles without finding a clear answer --I found everything and the opposite.

Should I graph the actual means as presented in the "descriptives" table, or should I graph the estimated marginal means?

  • 1
    $\begingroup$ Estimated marginal means are not only adjusted for the covariates; they are the predicted means with the collapsed subgroups equally weighted. For example, consider you have two factors with groups A1 A2 B1 B2 and you want to know EMMs for B1 and B2 groups. Then EMM for B1 is the simple mean of 2 means of predicted values: mean(mean_predicted_value_in_A1B1, mean_predicted_value_in_A2B1). $\endgroup$ – ttnphns Dec 20 '14 at 10:28

Report estimated marginal mean and standard error in writing results and making graph.


We're looking for long answers that provide some explanation and context. Don't just give a one-line answer; explain why your answer is right, ideally with citations. Answers that don't include explanations may be removed.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.