topic Re: Proc surveylogistic-strange value in Statistical Procedures

Proc surveylogistic-strange value

tinak — Thu, 11 Feb 2016 21:00:58 GMT

I get a strange value of >999.999 when I run proc surveylogisitic on a model. Anyone know why this might be?

Thank you!

Re: Proc surveylogistic-strange value

ballardw — Thu, 11 Feb 2016 21:05:07 GMT

"Strange value" for what? A model parameter, odds ratio, diagnostic or descriptive statistic.

Show the output with the value, but it is likely caused by data values.

Do you have some sort of value like 9999 that is used when something was missing or not collected?

Re: Proc surveylogistic-strange value

tinak — Thu, 11 Feb 2016 21:09:17 GMT

I should have been more clear for an Odds Ratio value.

Housing Homeless vs Apartment/house >999.999 <0.001 >999.999
##- Please type your reply above this line. Simple formatting, no
attachments. -##

Re: Proc surveylogistic-strange value

Reeza — Thu, 11 Feb 2016 21:46:39 GMT

Can you post your parameter estimates? I've seen this before and it has to do something with the parameter estimates but I need to jog my memory. Someone else may be able to see it right away.

Re: Proc surveylogistic-strange value

Rick_SAS — Fri, 12 Feb 2016 11:13:15 GMT

The odds ratio is printed with a special format, just like the more-familiar p-values in SAS. When a p-value gets to small, it is printed as "<.0001". In the same way, when an odds ratio becomes too big, it is printed as ">999.999 ". The true value is in the table (use ODS output to save the table to a data set), but you are seeing the formatted value.

If yuo want to know why your odds ratio is so large, that is a modeling question that requires knowing more about your model and data.