Erroneous Regression on Residuals
 

When I correct for the collinearity of regressors in a multivariate
regression by replacing with residuals from another regression, regression in
Excel is not working. One or more of the coefficients in the multivariate
system (incorrectly) go to 0, with a p-value of 1. When I have tried on other
machines, it runs fine. It seems like my stat pack is flawed. Has anyone else
encountered this? Is there a way to debug?

Scott -

You have not provided much information. What versions of Excel run fine, and
what versions don't? What "stat pack" are you using?

Excel's Regression tool in its Analysis ToolPak uses the LINEST worksheet
function for calculation. Its numerical properties have improved in recent
versions of Excel. See the Knowledge Base article:

Description of the LINEST function in Excel 2003 and in Excel 2004 for Mac
http://support.microsoft.com/default...b;en-us;828533

If that doesn't answer your question, please provide more information.

- Mike
www.mikemiddleton.com

"Scott" wrote in message
...
When I correct for the collinearity of regressors in a multivariate
regression by replacing with residuals from another regression, regression
in
Excel is not working. One or more of the coefficients in the multivariate
system (incorrectly) go to 0, with a p-value of 1. When I have tried on
other
machines, it runs fine. It seems like my stat pack is flawed. Has anyone
else
encountered this? Is there a way to debug?

I have had the same result, I used MINITab demo and got the correct output.
So there must be something wrong with Excel. I looked at the knowledge base
as suggested by Mike M., but a first glance it does not answer the question
why the the toolpak does not provide the expected outcome.

Below is my post:

I am using Data Analysis Toolpak|Regression to do a multiple regression
analysis of 3 data variables and 11 dummy variables (14 total). The ANOVA and
Statistics output is correct, but the coefficients for some dummy variables
and the intercept do not match the expected output.

Any thoughts?

I am using Excel 2003 SP1.


"Scott" wrote:

When I correct for the collinearity of regressors in a multivariate
regression by replacing with residuals from another regression, regression in
Excel is not working. One or more of the coefficients in the multivariate
system (incorrectly) go to 0, with a p-value of 1. When I have tried on other
machines, it runs fine. It seems like my stat pack is flawed. Has anyone else
encountered this? Is there a way to debug?

You are probably seeing the issue discussed in

http://groups-beta.google.com/group/...62751e581b6e26

Jerry

Keith wrote:

I have had the same result, I used MINITab demo and got the correct output.
So there must be something wrong with Excel. I looked at the knowledge base
as suggested by Mike M., but a first glance it does not answer the question
why the the toolpak does not provide the expected outcome.

Below is my post:

I am using Data Analysis Toolpak|Regression to do a multiple regression
analysis of 3 data variables and 11 dummy variables (14 total). The ANOVA and
Statistics output is correct, but the coefficients for some dummy variables
and the intercept do not match the expected output.

Any thoughts?

I am using Excel 2003 SP1.


"Scott" wrote:


When I correct for the collinearity of regressors in a multivariate
regression by replacing with residuals from another regression, regression in
Excel is not working. One or more of the coefficients in the multivariate
system (incorrectly) go to 0, with a p-value of 1. When I have tried on other
machines, it runs fine. It seems like my stat pack is flawed. Has anyone else
encountered this? Is there a way to debug?