For the fixed part - I would simply use the estimated fixed part terms - they are in the metric of the response variable. I typically draw graphs with predicted response on the vertical axis (same scale for all) and a separate graph for each predictor variable over the range of the predictor (deciles for continuous; 0 and 1 for dummies) and also plot (say) the 95% confidence interval around the partial line which I think is a helpful visualization about the comparative size of effect.
You can see the debate around this and appreciate why I do not like standardized regression coefficients
Thanks all for the kind suggestions, I've decided to go with Prof Jones's advise.
I've got another question regarding reporting of F statistics and contrasts statistics.
Hypothetically in a typical ANOVA, we'll write F(X,Y) = 11, where X is d.f of the effect and Y is d.f of the error.
In a mixed model, what is value Y if I'm going to report on fixed effect B in text - F(2,Y) = 11.355, p < 0.001. Also if interaction of A*B is significant, how do I report the statistics for its subsequent contrasts as well?
The model dimension of my data (from SPSS) is listed below.
I do not know how to do this in SPSS. But I think there is a solution in R, which uses the cohen's f2 as the effect size of a fixed effect. You can see this in the following link:
The effect sizes are estimated based on the Estimates of Covariance Parameters in the SPSS output. Variances between old/new models should be compared in the intercepts and here is the basic formula:
Variance in the old model - Variance in the new model / Variance in the old model.