Long time reader, first time poster. I'm doing an observational study and have gotten into unknown waters in both statistics and R (beginner in both).

(1) I have 2-categorical variable (species) and roughly 4000 rows with behavior (11 classes, reconfigured in to 'foraging' or 'not-foraging').

I'm trying to figure out if the two groups differ in their behavior My professor told me to use proportions of the behavior to do a t-test of the means of the two groups(species).

I've spend a week finding nothing but "you shouldn't do that" and "you can't do that" anywhere I look or ask online.

I'm at a loss for words - We're looking at basic university level, so I am not expected to figure out the secrets of the universe here.

I found prop.test, which I applied to the two groups, but I'm unsure of that actually shows, what I think it shows and I read somewhere that a prop.test is actually a z-test (I don't understand the difference).

(2) Furthermore I have another 3-categorical variable (vegetation type).

I'm trying to figure out if the behavior differs across vegetation type (never mind species for now - I can always do them seperately, if necessary). I was told I could do a simple, linear ANOVA type of model ( lm function in R), again; using the proportions of behavior as response and the vegetation type as predictor.

I've reached exactly the same dead-end as with question (1): Apparantly, it can't be done. I have found a few comment in different threads (present site included) stating that you CAN do a simple linear anova using percentage/proportion data, but I haven't found any actual examples of the procedure to do so and therefore I haven't gotten any further..

  • Have I misunderstood "Proportions"?
  • Is there a way of transforming my count-data into proportions and do a t-test?
  • Does anybody have any suggestions?
  • Is there any way of transforming my data into proportions and do an ANOVA?

I'm desperate and will appreciate any help/suggestions whatsoever.

  • $\begingroup$ I find it helpful to break down problems like these into what I want to predict (sounds like behavior) and what I want to use to make the prediction (sounds like you might have many, but species is the primary one). Can you put your variables into these two buckets? // The vanilla two-sample t-test works like this: some binary variable (control/treatment) predicts some outcome variable (e.g., height). $\endgroup$
    – Dave
    May 11 at 13:25

Your Answer

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

Browse other questions tagged or ask your own question.