Using solver for regression

**abhirules** · 04-22-2010, 12:12 PM

Hello folks,

I have a set of data (see attached excel sheet). I performed the generic multiple regression analysis to determine the x-coefficients. All of this is in the attachment.

My Problem:
As you may see i have a negative coefficient for one variable and a positive for another. I want to have a positive coefficient for both variables. How can i do that?

Thanks
AM

**shg** · 04-22-2010, 12:18 PM

What are you regressing against what?

Your title says you're using Solver, but there's no saved Solver model.

**abhirules** · 04-22-2010, 01:05 PM

Originally Posted by shg

What are you regressing against what?

Your title says you're using Solver, but there's no saved Solver model.

In my analysis,

Y = HBW4
X1 = BASIC
X2 = HIGH

Also, i have not used solver and that is what i am looking for. How can i use solver to limit the coefficients of X1 and X2 to be more than zero.

**shg** · 04-22-2010, 01:20 PM

See attached. The model is set up in Solver.

**abhirules** · 04-22-2010, 02:34 PM

Thanks a lot, i appreciate it

**koylita** · 10-27-2011, 04:00 PM

Originally Posted by abhirules

Thanks a lot, i appreciate it

The result is wrong because one of the coefficents goes to zero... if you do the same analysis with regression, you will get I

ntercepción 2357.339041
BASIC -2.012098045
HIGH 2.510001139

so the question is how to restrics the coefficients to be nonzero?

thanks

**shg** · 10-27-2011, 04:06 PM

Welcome to the forum.

The result is wrong because one of the coefficents goes to zero...

The OP's requirement was that all coefficients be non-negative, which yours aren't.

**koylita** · 10-27-2011, 04:34 PM

Sorry you are right... My problem is that... supose i need them to be different to zero...
my problem is that with simple regression i get the right coefficients but with multiple i always get one that goes to zero... and with the regression analysis all of them are diff to zero...

**shg** · 10-27-2011, 04:40 PM

Please start your own thread.

But before you do, be aware that you can't constrain a coefficient to be non-zero -- it makes no sense.

**koylita** · 10-27-2011, 05:03 PM

i would like to, but i don't have the button "new thread"... do i need an spefcial permission?
besides what i mean is that the solution in my case is wrong... is different to the one that goes with the add in "data analysis"