turn on suggestions

Auto-suggest helps you quickly narrow down your search results by suggesting possible matches as you type.

Showing results for

Find a Community

- Home
- /
- Analytics
- /
- Stat Procs
- /
- Need help with GENMOD, gamma distribution

Topic Options

- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Permalink
- Email to a Friend
- Report Inappropriate Content

03-12-2011 04:06 PM

I've learned how to use GENMOD assuming a poisson or negative binomial distribution, but I'm getting a much smaller deviance using a gamma with a log link. I need help interpreting these results.

I have a handful of explanatory variables, and the outptu is giving me the coefficient estimates for the intercept (1.87), and explanatory variables x1 (.29), x2 (.19), x3 (.47), and a scale parameter (3.87).

I'm trying to figure out where alpha is showing up on the output. The mean of the response variable is 7.13, so is alpha (7.13/3.87)? The results don't make the counts of my observed that well, but it's not bad.

Also, how do I find the corresponding model given these coefficients? I tried E(y)=exp(1.87 + .29x1 + .19x2 + .47*x3), but that didn't match my observed at all, and I'm not sure how to include the v and lambda variables.

I have a handful of explanatory variables, and the outptu is giving me the coefficient estimates for the intercept (1.87), and explanatory variables x1 (.29), x2 (.19), x3 (.47), and a scale parameter (3.87).

I'm trying to figure out where alpha is showing up on the output. The mean of the response variable is 7.13, so is alpha (7.13/3.87)? The results don't make the counts of my observed that well, but it's not bad.

Also, how do I find the corresponding model given these coefficients? I tried E(y)=exp(1.87 + .29x1 + .19x2 + .47*x3), but that didn't match my observed at all, and I'm not sure how to include the v and lambda variables.

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Permalink
- Email to a Friend
- Report Inappropriate Content

Posted in reply to deleted_user

03-14-2011 01:03 PM

Having specified a log link, you have indicated that E(Y)=exp(b0 + b1*x1 + b2*x2 + b3*x3). Whether you obtain a good fit or not is a different question.

I really don't know what you mean when you refer to "the v and lambda variables". GENMOD only allows you to model the expectation as a function of some predictor variables. It does not allow you to model the scale (or inverse scale) as a function of predictor variables. For that, you could use the NLMIXED procedure. But I am only guessing that your reference to "the v ... variables" refers to a set of variables which influence the scale parameter. And I have no idea what you mean by the lambda variables.

I really don't know what you mean when you refer to "the v and lambda variables". GENMOD only allows you to model the expectation as a function of some predictor variables. It does not allow you to model the scale (or inverse scale) as a function of predictor variables. For that, you could use the NLMIXED procedure. But I am only guessing that your reference to "the v ... variables" refers to a set of variables which influence the scale parameter. And I have no idea what you mean by the lambda variables.