On Sat, 20 Mar 2010, Bruno Thiago Tomio wrote: You can calculate R-squared in whatever manner you think might be appropriate for a random-effects model. There's no "correct" value in this case, but you could report the squared correlation between the dependent variable and the fitted values ($yhat), or do something with the sum of squared residuals ($ess). open greene14_1.gdt logs C Q PF panel l_C 0 l_Q l_PF LF --random R2 = corr(l_C, $yhat)^2 R2 = 1 - $ess/sst(l_C) Allin Cottrell