[R-sig-ME] proportions data

mikhail matz matz at mail.utexas.edu
Tue Nov 2 18:07:57 CET 2010

Hello colleagues,

I am looking at the functional correlates of fluorescent color in the  
coral larvae. My response variable ("redness") is a proportion,  
ranging from about 0.5 to 0.9. The distribution of these is not very  
normal, unfortunately.  Arcsin-sqrt  transformation does not fully  
solve the problem: about a quarter of all the groups that I am  
comparing still deviate from normality (lillie.test). The data tend  
to be ever so slightly skewed towards larger values, but with the  
sample sizes I have even a small deviation from bell-shape leads to  
rejection of normality.

  Is there anything I can do to be formally able to run mixed model  
analysis on such data?

Mikhail V. Matz
University of Texas at Austin

More information about the R-sig-mixed-models mailing list