Repeated measures ANOVA w/ time-varying covariate that does not vary across subjects

#1
Hello community:

I am seeking assistance with a repeated measures ANOVA. In this experiment, I examined the return rate of bees to colonies (= dependent variable). These colonies were subjected to one of four food treatments (= categorical predictor variable). The return rate of bees was evaluated once per week over a 13 week period. Thus, in this analysis "colony" is the subject (nested within treatment) and "week" accounts for the repeated measure.

I am testing the effect of food treatment, but at the same time, I would like to assess the effect of colony size and the local abundance of flowers on the return rate. Both are time-varying covariates; however, flower abundance is a weekly measurement that *does not vary across subjects/colonies.*

Is it possible to analyze the effect of flower abundance given that each week corresponds to a given value of floral abundance? I've tried coding it as a normal covariate, and SAS returns no result for this variable. Time is significant in the model, but having graphed out the data, I am confident that this is due to changes in floral resources and not time itself.

Any suggestions for how to move forward would be greatly appreciated!
 
Last edited:

MBJ

New Member
#2
Re: Repeated measures ANOVA w/ time-varying covariate that does not vary across subje

Which procedure do you use?

Morten
 

MBJ

New Member
#4
Re: Repeated measures ANOVA w/ time-varying covariate that does not vary across subje

Could you use flower abundance as a continous variable instead of your time variable?

Morten
 
#5
Re: Repeated measures ANOVA w/ time-varying covariate that does not vary across subje

I'm not sure I understand. Floral abundance, when I include it, is a continuous predictor variable (covariate). The week is my time measurement. Are you suggesting that I use floral abundance instead of "week" as my repeated measure?
 

MBJ

New Member
#6
Re: Repeated measures ANOVA w/ time-varying covariate that does not vary across subje

It depends on how your model is set up. My suggestion was to exclude your time variable in the model statement. If you have colonies as random effect you will still have a repeated setup although without a time element. If you want to control the correlation within subjects in time I think you still can have your time variable in a repeated statement.

Morten