Dear all,
I'm intended to model the effects of a number of categorical factors (season, sex, age) on the proportion of lean birds using a beta regression (R-package 'betareg'). The dependent variable had 0's and 1's, so I proceeded to transform the data using the formula (y * (n-1) + 0.5/n) where n is sample size (Smithson and Verkuilen 2006). I've tried running the model, but the error message stating that dependent variable is invalid because all observations must be in (0,1) keeps popping out. I checked out the transformed variable and 0.00 observations were indeed transformed into 0.02083333. Any suggestions on how to get around this issue?
Thanks,
I solved it by eliminating the outliers of the dependent variable.
Hi Henning Teickner!
Thanks for your suggestions. I'm gonna double-check the typos entered in the rescaling formula which I hadn't considered before. After having had a bit of struggle to getting the transformed data ready for analysis, I did a brief search for R-packages featuring beta regression models that could handle data values with zero and one, and I think I may have downloaded and skimmed thru "zoib".
I realized that they use the bayesian approach which I'm complete unfamiliar with. I kind of "chickened out" and persisted in finding ways to make my transformed data workable despite the fact Im not very fond of data transformations. So, I guess this may be an opportunity to venture out into uncharted learning territories. Thanks again for your valuable comments.
Best regards,
Alain
- Badges
- Science topic
- Similar topics
- Biological Science
- Faunistics
- Ornithology
- birds