Dear all,
I'm currently conducting the tests for my masters trying to analyze the outcome from regression results.
Unfortunately, no matter what I try the regression coefficient always shows very small values ( .004; 4.286E-5) while the estimations created in Excel show more valid results.
My variables contain binary values (0,1) (this is also the DV), percentages and scales. Could that have anything to do with my issues?
Many thanks
If you are using an OLS package to do you regression that is probably not what you want to do. See this link for aid:
https://www.google.com/search?q=regression+on+percentages&rlz=1C1CHBF_enUS847US847&oq=regression+on+percentages&aqs=chrome..69i57j0l3.14819j0j8&sourceid=chrome&ie=UTF-8
Best, David Booth
Hello Lars,
The unstandardized regression coefficients depend on the scale/s chosen for the respective independent variable (IV) and dependent variable (DV). Change the scale, and the values will be altered. Standardized regression coefficients should be constant, regardless of linear transforms of the respective IV and/or DV scales. I'm not sure what comparisons you're running in Excel (bivariate regression only, or MLR), but whenever you look at an ensemble of predictors, the resultant estimates of regression coefficients might not match well to single predictor (bivariate regression) model results, due to collinearity and/or possible suppression effects.
Since your query isn't really clear as to the nature of the set of variables you're working with, it would be wise to check the source offered by David Eugene Booth .
Good luck with your work!
Hi Lars,
the nature of your IVs and DVs determine the estimator that you would use..
Since your question is not that clear, kindly redraft..
Cheers
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software
- SPSS