Hi,
I am a co-author of paper where we used a General Additive Mixed Model to model the relationship between body length and trophic position of aquatic animals.
I just received a comment from a referee that the histogram of the model residuals is not very informative, and we should provide a q-q plot for random factors and a plot of the residuals vs fitted values.
I do understand the need for the residulas vs fitted values. A trend in this plot would be indicative of a problem.
However, I do not know how to interpret the q-q plot for the random factors. I know what a q-q- plot represent, but when I ask to plot the GAMM model (using R), the first plots are plots of effect-Gaussian quantile for each random factor (I think this is the plot that the referee is talking about).
What does this effect represent? Having points that do not follow the straight line, extreme points very positive or negative for example, is a problem for the analysis? Or is just an indication that some levels of the random factor have more positive or negative relationships?
My guess is that the q-q plot for random factors are the effect-quantile plot, but I do not know what the effect stands for and if a deviation from the line would represent a problem or not.
Thanks in advance for any advice.
Take a look at the sjPlot and sjstats packages, they have excellent diagnostic plots for gam
- Badges
- Science topic
- Similar topics
- Cognitive Science
- Thinking