# [SOLVED] Regression of many variables

• Jul 31st 2009, 10:22 AM
Nialsh
[SOLVED] Regression of many variables
I have a pretty good understanding of regressions from high school statistics, with one independent variable and one dependent variable.

I would like to create a regression with a vector of N independent variables and a vector of M dependent variables. At this point I don't know what shape my data will be, so I figure a high-order polynomial regression will work. Maybe something of this form:
$y_i = \sum_{j=1}^N\sum_{k=0}^{order}{a_{ijk}*{x_j}^k}$

Is this something that is done very often or am I on the wrong track? I would be happy to just get some links.

Thanks,
Neal
• Jul 31st 2009, 09:39 PM
matheagle
By letting $w_{jk}=x_j^k$ this is linear in the coefficient, a's.

So you can apply multivariate regression here.
The question is, whether or not your errors are normal or not.

See 'General linear data model' at http://en.wikipedia.org/wiki/Regression_analysis
• Aug 5th 2009, 09:03 AM
Nialsh
Cool, thanks for that. I did a lot of reading and it turns out I was looking for supervised learning.