I have data of relative abundance (in percentage) of 5 categories of ingested items in the stomachs of 8 fish species collected from 4 rivers. The number of sampled fish stomachs collected from different rivers varied from 0 to 12 and not all stomachs contained all 5 categories of ingested items. Can I use GLM to test if there's significant difference of relative abundance of each ingested item among the two main effects: species (8 levels) and river (4 level) with no interaction? Is it correct that I cannot test the interaction effect of species and site because some combinations are missing from the data (ie. I collected 0 sample for fish species 1 from river A). And does the small sample size affect the results? I used the following codes and the KS test from DHARMa package gave non-significant p. value :
Food1_mod<-glmmTMB(data=Stomach, Food1 ~ Species + River), family="beta_family")
glmmTMB:::Anova.glmmTMB(Food1_mod)
Response: Food1_mod
Chisq Df Pr(>Chisq)
Species 29.852 7 0.0001011 ***
River 12.932 3 0.0047863 **
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
Thank you!