Two-way partially crossed design - model selection and inference


New Member
I have data from an ecological experiment featuring a partially crossed design. The experiment manipulated air temperature (3 levels: T1, T2, T3) and precipitation (2 levels: W1, W2). However, the T2-W2 treatment combination was not performed. We have a number of response variables (e.g. plant biomass, soil Carbon, etc.), measured once. I did not see any evidence of interactive effects in the data, thus I initially analyzed the data with one-way ANOVA's. A reviewer has made it clear that they would prefer to see the data analyzed with a two-way ANOVA for the treatment levels that are crossed. Previous efforts using this same experiment accommodated this by using a two-way ANOVA for the fully crossed temperature and precipitation levels (T1-W1,T1-W2,T3-W1,T3-W2), and a one-way ANOVA on temperature only (T1-W1, T2-W1, T3-W1, T3-W2) . I found this made for clunky reading, though, and I'm trying to figure out if there's a single model structure that can both examine the individual and interactive effects of temperature and precipitation, while also allowing for post-hoc means separation for all levels of temperature as well as the crossed levels of temperature*precipitation. Thank you in advance for your time!


Active Member
You could try a GLM. Usually you can nominate the interactions you want.
Second thoughts, probably not, because there would be missing data.
Last edited: