topic Can SAS 9.1 calculate the Information Value? in Statistical Procedures

Can SAS 9.1 calculate the Information Value?

deleted_user — Fri, 09 Jul 2010 17:39:39 GMT

Hi all,

Please do you how to get SAS to compute the Information Value (for categorial analysis)?

PROC FREQ data=work.dataset;
TABLE SuccessfulSale*Occupation / chisq cmh;
run;

Syntax above compute the chi-square statistics to help user study whether a successful sale (SuccessfulSale = 1) is related to a customer's occupation (11 categories of occupation).

But this doesn't produce the information value (IV).
Grateful if you could shed some light on how this could be done via SAS 9.1

Many thanks in advance.
Regards
Silverflute

Re: Can SAS 9.1 calculate the Information Value?

Dale — Fri, 09 Jul 2010 18:45:00 GMT

I have not heard of any "Information Value" statistic. I have heard of Information Criteria statistics. There are a number of Information Criteria statistics

    - Akaike Information Criterion
    - Schwarz Bayesian Information Criterion
    - Hannan-Quinn Information Criterion

There are probably other Information Criteria statistics. For the statistics mentioned above, there can be a number of variants of these Information Criteria in different settings.

Are you looking for one of these Information Criteria statistics?

Re: Can SAS 9.1 calculate the Information Value?

deleted_user — Mon, 12 Jul 2010 14:57:33 GMT

Information Value (IV) measures the predictive power of a particular variable.
I heard there is a PROC VARCLUS but I don't know how to use this to generate the WOE (Weight of Expectation) and IV for a particular variable X on it's abilty to predict variable Y. I found the formulae that relates WOE and IV from the % successful and % unsuccessful, and since my boss need the results asap so I'll manually compute it from excel.

Grateful if anyone can shed some light on whether there is a SAS command that can list out the Information Value of a particular variable X (Low, Medium, High) to predict a particular dichotomous variable Y (Success =1, Unsuccessful =0).