ALL POSSIBLE COMBINATION WITH INDEPENDENT VARIABLE

[caetano]

Hello everyone,

I need to do a probit regression with 5 independent variables but I have 16. In other words, I need to do all possible regressions with 16 variables 5 at the time - something like this:

Y = a + x1 + x2 + x3 + x4 + x5
y = a + x1 + x11 + x13 + x14 + x6
y = a + x10 + x2 + x9 + x8 + x7
 ...

In the end I need the expression with the highest R^2 and the independent variables should be independent and with the lowest AIC information.

Is there any way to perform this?

Thanks for the help,
João Caetano

[David Marso]

Check the archives of this group on how to generate combinations.
Have you thought about generating them with nested loops to create a file of variable indices then use these indices to build your command syntax?  Probably SWIM will come up with a Python solution ;-)
--

caetano wrote

Hello everyone,

I need to do a probit regression with 5 independent variables but I have 16. In other words, I need to do all possible regressions with 16 variables 5 at the time - something like this:

Y = a + x1 + x2 + x3 + x4 + x5
y = a + x1 + x11 + x13 + x14 + x6
y = a + x10 + x2 + x9 + x8 + x7
 ...

In the end I need the expression with the highest R^2 and the independent variables should be independent and with the lowest AIC information.

Is there any way to perform this?

Thanks for the help,
João Caetano

[Bruce Weaver]

This sounds like a fishing expedition to me.  Have you ever looked at Frank Harrell's book on regression models?  Mike Babyak's nice article summarizes the main points about over-fitting.  You can get it here:

  http://os1.amc.nl/mediawiki/images/Babyak_-_overfitting.pdf
  http://www.psychosomaticmedicine.org/content/66/3/411.abstract

HTH.

caetano wrote

Hello everyone,

I need to do a probit regression with 5 independent variables but I have 16. In other words, I need to do all possible regressions with 16 variables 5 at the time - something like this:

Y = a + x1 + x2 + x3 + x4 + x5
y = a + x1 + x11 + x13 + x14 + x6
y = a + x10 + x2 + x9 + x8 + x7
 ...

In the end I need the expression with the highest R^2 and the independent variables should be independent and with the lowest AIC information.

Is there any way to perform this?

Thanks for the help,
João Caetano