Dear list, Inspecting residuals of my linear models, I detected spatial autocorrelation. In order to take this into account, I decided to use the GLS method with the correlation = corGaus ( ~ X + Y).
Then, I can sort my GLS models based on their AIC. But ... how to know the proportion of the variance explained by the best one (it can be best of the worst models) ? R-squared value has not the same meaning for OLS and GLS ... - Could the R2 value calculated with the OLS model (using lm) constitute a potential proxy of the variance explained by the GLS model ? (the answer is probably no) - Is a R-squared based on sqrt(cor(obs, predicted)) a better approach ? - What about pseudo R-squared like Nagelkerke's ? Suggestions for any better approach are welcome ! Thanks in advance, Arnaud ______________________________________________ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
