1

I want to compute a pseudo-$R^2$ for a model whose parameter estimation was based on maximum likelihood (function likfit(), package geoR, R software).

I tried to compute the $R^2$ proposed by Maddala (1983) which compares the maximized likelihood for the model without any predictor and the maximized likelihood for the model with all predictors. I got a really low value (0.01%). Did I miss something? Are there other $R^2$s which are more appropriate than the $R^2$ of Maddala?

gung - Reinstate Monica
  • 132,789
  • 81
  • 357
  • 650
Jacquot Marion
  • 11
  • 2
  • 1
    Without more information, it will be difficult for people to say whether the low value is reasonable or not. Could you explain how you computed the $R^2$? Can you provide a simple, short set of sample data exhibiting the apparent problem? – whuber Nov 07 '12 at 19:16
  • 1
    You may find the [pseudo-$R^2$ page from the UCLA stat consulting group](http://www.ats.ucla.edu/stat/mult_pkg/faq/general/Psuedo_RSquareds.htm) helpful. You may also find this CV question: [Logistic regression: Which pseudo R squared measure is the one to report](http://stats.stackexchange.com/questions/3559/) worth reading. – gung - Reinstate Monica Nov 07 '12 at 19:43

0 Answers0