I have one dependent variable name as "win ration" of the deal contested and more than 30 independent variables, all are categorical variable name as role of the customer, geo, region, and 27 competencies marks between 1 to 12 for each competency (this is like the performance parameter "1" is high and "12" is low rating) name as comp1,comp2,....,comp27

My question is that which is the best model or predictor model to filter out which all competencies and other variables are really affection the win rate.

for this i used beta regression but non of the competencies are coming out significant and when i perform the step wise for dim reduction, this method is not working on beta reg. In this case all competencies are treated as quantitative variable, the reason is that all are bound between 1 to 12 and if i will categorize this all 27 competencies will have 12 categories

Please help me to do this analysis

  • $\begingroup$ Start with simple univariate analysis to identify if you have any relationships at all to begin with. $\endgroup$ – Arun Jose Aug 11 '14 at 11:36
  • 1
    $\begingroup$ Univariate analysis is not usually helpful and is often misleading. $\endgroup$ – Frank Harrell Aug 11 '14 at 11:48

The data are not usually capable of telling you what the variables that "really" affect Y are. It would be a good approach to use data reduction techniques (variable clustering, etc.) to reduce the number of potential predictors by grouping them so that collinear predictors are not competing against one another. You can use a smaller number of cluster scores (e.g., first principal components) as predictors in a smaller model. Don't try to remove variables.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.