Hello, I have a dataset of floating leaf asymmetry indexes and would like to run a GLM using this index as a response variable and the other characteristics of the leaf as predictor variables (leaf width, leaf thickness).
the variable response is the index ranging from 0 to 0.83. that is, a continuous variable.
I can not use the gamma family for having 0 values, I can not use the negative binomial family because the values are not integers. Would anyone know which family is the best?
m1<-glm.nb(IND~ALT+ESP+FORM+LOCAL+FILO+AREA, link="log",data = amostra)
IND = asymmetry index
LOCAL = Forest A or Forest B
ALT = height of the tree (continued)
ESP = sheet thickness (continuous)
FORM = leaf shape (categorical)
FILO = phylloxia (categorical)
AREA = leaf area (continuous)
Does it make sense for me to transform the response variable in this case? If I use the square root of the index the distribution goes from 0 to 1 (as in the image)