Continuous versus Ordinal effects in Logistic Regression
Jul 22, 2014 10:58 AM(1133 views)
I have some continuous effects in a Logistic Regression problem. Is there any reason why the regression results would be better or worse if the continuous variable is binned into a small set of ordinal variables? If it is left as a continuous variable, then there needs to be just one coefficient. So, it would seem that leaving it as is will be better, since the regression has to estimate just one coefficient. Could there be any benefit to breaking up the continuous variable into a small number of ordinal groups?