I have a fixed Factor with 8 levels (Condition), alongside two fixed effects (two predictor variables) in a linear mixed-model. I would like to compare the total effect of condition on the dependent variable, alongside the effects of the two predictor variables.

In R, the output of summary(my.model) gives the following headings, however it splits the Condition into 8 separate estimates instead of giving one beta-coefficient for the effect of Condition.

Fixed effects:

        Estimate Std. Error df t value Pr(>|t|)

I know to obtain the F-statistic for Condition I can do a likelihood ratio test against the null model and report both the parameter table and the effects table. However, I would like to make my results report as concise as possible. My question is whether (in general) reporting just the main-effects table is enough in describing the results of a mixed effects model, or whether there is a way to include the total effect of Condition in the parameter table.

  • $\begingroup$ Do you have idea what total (overall) effect is? $\endgroup$ – user158565 Nov 26 '18 at 18:13
  • $\begingroup$ You need to explicitly define what you mean by the total effect of Condition. $\endgroup$ – Dimitris Rizopoulos Nov 26 '18 at 19:40
  • $\begingroup$ For the total effect using the LRT, I get χ2(7)= 218 when I compare the model with Condition to that without. I am looking for something equivalent to a beta-coefficient (Estimate in the R summary output) for the effect of condition across all levels of Condition. Some parameter that will allow me to directly compare the influence Condition (across all level) on the dependent variable with the influence of the fixed predictors. $\endgroup$ – Cipriana Nov 27 '18 at 14:21
  • $\begingroup$ I think after looking into it more, I was over-complicating reporting the results. There doesn't seem to be any standard way that the results of LMM must be presented so I think as it is the fixed effects I am interested in, I can just report the main-effects table with F-statistics (with Kenward-Roger approximation). This allows me to report the overall variance explained by Condition and to omit the parameter table to keep the results concise. $\endgroup$ – Cipriana Nov 30 '18 at 11:53

Your Answer

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

Browse other questions tagged or ask your own question.