linear regression and results

felixxxland · August 24th, 2004, 3:38 pm

I posted this question in an other forum, but i think that it will rather be answered in the students forum:i am doing a project on linear regression-style analysis. my question:is it possible/reasonable to have negative R squared values in the regression result, especially when i have side constraints such as the regression weights have to sum up to 1. could you also name academic references plse?i do get them and want to be sure that it is not a programming mistake.thank you allf.

jamiepiolin · August 24th, 2004, 3:57 pm

There should be no negatives in your r-squared. Perhaps you're blowing up the regression by having multicollinearity among your indexes. Bill Sharpe's article in the Journal of Portfolio Management in Winter 1992 should help.

Droplet · August 24th, 2004, 6:59 pm

negative can be only adjusted R-squared, it happens sometimes.. specially in corporate financeJust R-sq is always >=0DROPlet

JamesH83 · August 25th, 2004, 6:09 am

hahaha you have obtained a complex correlation coefficient how interesting