on rule #3: in other words, it's not enough for a variable, in order to be included in the model to avoid bias, to be CORRELATED with the dependent variable even after controlling for variables already included - it must AFFECT the dependent variable? the two seem not the same. e.g. in our model (regime~oil), islam is correlated with regime (and w/ oil) even after controlling for other vars, but it probably doesn't "affect" regime causally, or there's at least no cohesive theory that it does. should it be included? if yes, then "affect" is only meant as correlation, right? omitted variable bias occurs UNLESS one of those three are not true. also change #3 to "it affects the dependent variable even fater controlling for the other variables already in the regression" Gary On Sun, 11 May 2003, Olivia Lau wrote: _______________________________________________ gov2001-l mailing list gov2001-l(a)fas.harvard.edu http://www.fas.harvard.edu/mailman/listinfo/gov2001-l _______________________________________________ gov2001-l mailing list gov2001-l(a)fas.harvard.edu http://www.fas.harvard.edu/mailman/listinfo/gov2001-l