Serving the Quantitative Finance Community

 
User avatar
TedSpread
Topic Author
Posts: 0
Joined: October 12th, 2005, 5:23 pm

multi variate regression

May 1st, 2006, 6:36 pm

how easy (and quick) is it to run a mult-variate regression with one dependent and 100 independents? where can this be done (matlab?)thankst
 
User avatar
Lepperbe
Posts: 0
Joined: November 9th, 2002, 10:36 pm

multi variate regression

May 1st, 2006, 7:18 pm

you forgot to ask how useful...
 
User avatar
gardener3
Posts: 8
Joined: April 5th, 2004, 3:25 pm

multi variate regression

May 1st, 2006, 7:28 pm

If this is for a financial or an economic application, and if you have 100 idep var, you are probably doing something wrong (unless you've got panel data). It is easy in matlab: inv(X'*X)*X'Y. X is matrix of indep vars, Y is a vector of dep vars.
 
User avatar
bostonquant
Posts: 0
Joined: March 26th, 2006, 7:46 pm

multi variate regression

May 1st, 2006, 7:56 pm

When you overfit a model you are adding terms that contribute little, hence you are adding variance as compared to the simpler reduced model. There is no way you will have 100 significant independent variables. The bond market can be explained with a handful of regressors and the equity market with depending upon what you are doing over 10.Look at MSE with the full model and look at MSE with a "reduced model" and you will see what I am saying. Stepwise and backward/forward tecniques don't always give the best results either so be careful.