HOW TO CHOOSE EFFECTIVE PARAMETERs IN STEPWISE REGRESSION?

Dear friends, I got 80 independent variables and one dependent variable with 1000 datasets. I was performing stepwise regression with quadratic fitting. MATLAB memory is showing an error of "Out of Memory" in very starting of its execution. Definitely it is because of large number of variables. Now I am planning to select important (significant) variables first (out of 80 variables) to run my program. Could you please help me how to execute this task.

Answer 1

[the cyclist]

I'm not so sure that 1000 observations of 80 variables should cause an out-of-memory error. I just successfully ran the code

N = 1000;
X = randn(N,80);
Y = randn(N,1);
[b,se,pval,inmodel,stats,nextstep,history] = stepwisefit(X,Y)

Are you sure there is not some problem with the setup? Can you post your X/Y data?

I suppose one way to test which independent variables are more likely to be important in predicting the dependent variable is to calculate the correlation coefficients between them.

[Ajay Goyal]

Thank You very much Sir for your valuable reply. I have learned a lot from your posted answers. I will try to implement your idea to find covariance coefficient matrix.Data I am working on is highly confidential and hence can not be communicated. My variables are like A,B,C,AB,AC,BC,ABC i.e. I have tried will all possible combinations. My idea behind using all possible terms is that model first finds all significant data and after removing outliers, it gives a best fit with overall adjusted R2 greater than 0.9. Sir, please let me know that weather I also need to calculate covariance of interaction terms. Thank You once again.