I am studying the effects of Habitat (flower diversity) on male/female bee abundance. But I can't really get my head around the statistical approach.
I want to check if there is a difference in abundance of male and female bees, between different habitat types.
Bee male and females are counted in multiple habitats, in three different rounds in the year. For each habitat, an inconsistent amount of transects (samples) (ranging from 10 to 24) are walked and bees are counted.
I thought of using a Generalized Linear Mixed Model for this (because of the repeated survey rounds and possible interaction between males and females), but am not sure how to use it in this case.
Do I need to make a male-female ratio for this analysis? I thought it was better to keep the males and females separated, because then I could test for interaction between them.
What should I do with the different amounts of transects? Can the sample size per habitat be different?
I hope someone can help me out.
Thanks in advance,
Rick
I think one-way ANOVA should do the trick. Or at most two-way ANOVA, depends on how you arrange your datasets and your objective. After you done the ANOVA, proceed to post-hoc test to see which groups has significant different.
Thank you! I was thinking about that, but then I can't correct for the 3 multiple rounds (which are pseudoreplicates, aren't they?). I can carry out ANOVA's for the separate rounds, but that's not quite what I want.
You can answer your basic question with a chi-square of independence between habitat and proportions of male/female bees. You can do this over all rounds or per round. See the R script in attachment.
For post hoc tests, see the R companion:
https://rcompanion.org/rcompanion/b_05.html
If you want to test whether the sex ratios are different between habitat types, you have a few options.
I suppose you have a dataset like this:
Habitat Transect Sex.ratio (f/(m+f)) Bee.abundance Season
Forest 1 0.44 43 Spring
Forest 1 0.47 32 Summer
Forest 1 0.45 52 Fall
Forest 2 0.48 39 Spring
… … … … …
Field 41 0.56 24 Spring
Field 41 0.51 19 Summer
… … … … …
With Season being a factorial variable with the 3 times you repeated the measurement (e.g. Spring, summer, fall). Each group of 3 site visits is nested in a certain transect, so they are pseudoreplicates. To account for this, you can use mixed modelling with ‘Transect’ as a random variable to account for variation between sites.
Probably your best option is to use a generalized linear mixed model, in which each individual bee is modelled as a 0 or 1 (for male and female respectively). You could model each bee individual separately, if your dataset is arranged so that each individual is a single row. However, this would be a rather big dataset. On the other hand, total bee abundance per site visited in combination with sex ratio is as informative, as sex ratio*Bee abundance = total number of females. R-code for this would be:
library(lme4)
beemodel
- Badges
- Science topic
- Similar topics
- Biological Science
- Physiology
- Physiological Processes
- Stress