I am giving different doses of a drug to mice and measuring a quantitative response. I want to determine if there is a linear relationship between the dose of the drug and the response. Pearson correlation seems to be able to do that.

But I am testing multiple mice per each dose, so I have several observations of the dependent variable per each level of the independent variable. And I am interested in the correlation between the dose and the true population mean of the response in each dose. I would like to know the r and p values (in standard Pearson terminology), i.e. how strong the relationship is and how statistically significant it is.

So I need something like Pearson correlation on repeated measures, but I wasn't able to find any application of that (although I thought it is a pretty typical situation and common analysis). Could you please suggest how to do such an analysis?

  • $\begingroup$ Why did you say of repeated measures? I don't see it here. It sounds to me from your description that you a kind of give dose I to mice 1,2,3; dose II to mice 4,5,6, etc. $\endgroup$
    – ttnphns
    May 30, 2014 at 20:22
  • $\begingroup$ Yes, I had that design in mind. "Repeated measures" might be an incorrect description. I though that maybe we can think of it as "repeated measures" of the same dose, but I am not sure if that is valid. At the same time, I would be also interested in how to analyze the true "repeated measures" design, in which each mouse receives each dose. If there is a solution to that, it would also work for me. $\endgroup$
    – Viktor
    Jun 2, 2014 at 17:55


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.