[Home Page] [Previous Page]
[Old Roundtable Archives] [New Roundtable]
Posted by rcolberg on January 11, 2000 at 02:09:38:
In Reply to: Re: Negative Beta Values posted by Lisa on January 10, 2000 at 19:12:02:
First, I would check for multicollinearity or correlation between your "independent" variables. If you have a relatively small sample multicollinearity can cause problems like this. Look at the correlation matrix between independent variables, always a good idea in small sample regressions.
You can run a regression with only the seat comfort variable and observe the sign and significance--this should help.
Finally, whether the seat comfort should be in the equation can be determined by running the regression with only color of car and gas mileage as independent variables. Note the adjusted R square and compare it to the adjusted R square for your model. If there is little difference, or if your model has a lower adjusted R square ignore seat comfort it contributes nothing. The lower beta (assuming about the same variance in each variable) suggests it is less significant than the others.
Interesting problem, hope you find the answer.
Subject: