Statistical test for categorical variable and two numerical variables?

Hi there, I'm hoping to get some help on which statistical test would be appropriate for the data I'm trying to sort out. I am looking at the abundances of birds which are either "Migrants" or "Residents" over three time periods in the spring. I'm hoping to see if there is a significant difference between the number of Migrants and the number of Residents, as a function of the time period. My data are as such (for example):

Time Period Migrant Count Resident Count
Early 1 3
Early 2 4
Mid 0 2
Mid 3

Welcome to the community!

Since your problem is to check whether three is a significant difference between the number or Migrants and the number of Residents over different time periods of spring, I think you may do ANOVA here, provided the assumptions hold good.

You can compute Migrant_Resident_count_difference as the difference of Migrant Count and Resident Count. Then, you can do ANOVA on these differences with respect to Time Period.

Hope this helps.


I should emphasise that though ANOVA can be used, it's debatable whether you should use it or not. You've count data, so surely the assumption of normality will fail. If there are "large" number of observations, then data can be assumed to have come from a Normal distribution (saisfying conditions of Normal approximation to Poisson / Negative Binomial), and then you can use it, otherwise you should not.

Thank you for the input. You're right, my data are likely not normally distributed. Considering the other stats I've calculated don't seem to hold any significance either, I may forgo this calculation and just present what I've gotten so far.


This topic was automatically closed 21 days after the last reply. New replies are no longer allowed.