[R-sig-eco] Q: variable selection prior to Path analysis (lavaan)?

[Adnan Sisic]

Dear Sig. Ecology Members,

I have a data set which contains frequencies of occurrence of 8 different species and >35 environmental variables. I want to determine main environmental drivers of each species occurrence and considering path analysis. I would like to avoid including all of the env variables in path model (tried, some obvious effects turn out non significant) thus would like to ask:
Is it correct to perform step-wise selection and build path model by including only the most influential variables for each species based on the stepwise selection results? If not, is there any other way to select variables prior to the path model; alternative data analysis suggestion are welcome (other than rda and similar)?
Thanks in advance,
Adnan


-

	[[alternative HTML version deleted]]