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
- /
- SAS Programming
- /
- General Programming
- /
- Regression output

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
- Highlight
- Email to a Friend
- Report Inappropriate Content

11-24-2016 09:18 AM

Hello everyone,

I am working on a Zero-Inflated Poisson model (see program below). For one of the variables (n1), SAS calculates its estimate but not its standard error, nor t-value, nor p-value. Instead of the respective value, the SAS output shows a dot (see file attached).

Does anyone knows what it means and and why it is not computing the different values?

proc countreg data = data.wtovxc method = qn;

model y = n1 unesco_avg polity_avg/ dist= zip;

zeromodel y ~ n1 unesco_avg polity_avg;

run;

Thanks a lot.

Jeanne

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

11-24-2016 07:46 PM

n1 (and inf_n1) both have no standard error, t, or p. But they both also have are reported to have zero degree of freedom. That's probably the revealing bit of information. I'm not familiar with COUNTREG, but in other countexts a zero DF suggest no variability of the variable in question.

So, I'd suggest first looking at the distriubtion of n1 and inf_n1 for all the records that were accepted by your proc countreg.

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

11-25-2016 04:58 AM

Thanks for your reply. I have checked the distribution of n1 and there is some variability. I did not check inf_n1 since it is build by the regression from n1. So it is the same variable.

I have also tried the regression with the same variables using Poisson and Zero-Inflated Poisson models, which work. However, when I try using negative binomial or zero-inflated binomial models, I get this result (zero degree of freedom, no standard error, t, or p).