1
$\begingroup$

I have a data set of a 3-way mate choice experiment between species a, b and c. I aim to test whether males courted more often with females of their own species. The dependent variable is count data of number of courtships with females of each species. This is what my data looks like:

enter image description here

Does anyone know an appropriate statistical test to use in this case?

$\endgroup$
  • $\begingroup$ Looks like you need a Poisson model. $\endgroup$ – user2974951 Jul 9 '19 at 11:08
  • $\begingroup$ Would you like to test whether individuals are more likely to mate within their own speacies or also if that's more common for a given species of the male? $\endgroup$ – David Jul 9 '19 at 11:16
  • $\begingroup$ So it would likely be a Poisson regression or negative binomial regression? Thanks, David - initially I want to test whether individuals are more likely to mate within their own species, but then afterwards if it is more common for a given species of male. $\endgroup$ – VRWizard Jul 9 '19 at 14:06
0
$\begingroup$

Your data is not currently in the correct shape for a poisson/negative binomial regression because you currently have three columns that constitute your response variabl. As such, you will need to reshape your data to answer the question you are interested in.

I think you are actually after a chi-square test but this will require that you reshape your data into a two by two contingency table.

If you want to account for the potential effect of an individual (I.e. a male that is particularly keen on a particular species) you might extend this to be a log-linear model with mixed-effects. You can do the latter in R with the lme package.

Note: A log-linear model is a specific case of a poisson/negative binomial model, and you almost always want negative binomial errors.

EDIT: enter image description here

A table like this (albeit editted to suit your stats programs needs) would yield the results I think you want by using a chi-square analysis. You wouldn't be combining counts between species so you shouldn't run the risk of doubling as you suggested in the comments.

| cite | improve this answer | |
$\endgroup$
  • $\begingroup$ Thanks André. I'm not sure how I can reshape the data to a 2 by 2 contingency table without affecting the overall analysis. If I were to combine the courtship events of the two unrelated species to one column, it wouldn't be an equal comparison as the number of females encountered would be double the number of females from the same species. Is this something I could incorporate into the model? $\endgroup$ – VRWizard Jul 9 '19 at 16:53
  • $\begingroup$ Could I do multiple pairwise comparisons of the number of courtship events between females of two species whilst only examining one male species at a time? For instance, the first comparison would be all males of species a, with the dependent variables being female species a and female species b. Could a glm with poisson or negative binomial be used here or something else? $\endgroup$ – VRWizard Jul 10 '19 at 13:05
  • $\begingroup$ Sorry for the long delay - I have been away on vacation and have just gotten back to work. I have updated my answer to display what your contingency table might look like. $\endgroup$ – André.B Aug 4 '19 at 21:16

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.