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
- /
- Negative binomial output: p = 1.000?

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

11-06-2014 07:03 PM

Hi all,

This is probably a bit of a silly question, but I can't seem to find anything about this. I know in SPSS with other tests, a p value of 1.00 would mean absolute certainty and there is usually something that isn't quite right, but I'm not sure what a p value of 1 means in the output for negative binomial or if it is different in SAS. I have a two variables, plus the interaction between the two (which is what I'm interested in). Can anyone help me so I know how I should be interpreting this result or if there is something that I possibly did wrong that I need to redo?

Thanks

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

11-07-2014 10:22 AM

If you are saying that you fit a negative binomial model including interaction, like this:

proc genmod;

model y = x1 x2 x1*x2 / dist=negbin;

run;

and that the p-value reported for the x1*x2 interaction was 1.0000, then this simply means that there is no evidence that the interaction parameter is nonzero.

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

11-07-2014 10:30 AM

If you have a database with lots of lots of observations the p-value (excuse my language) "don't mean shit".

It's going to be 1 no matter what you do. So in this particular case you'll have to find another type of measurement

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

11-07-2014 10:56 AM

Hmm...

Too much data ==> Meaningless results

Hmm....

I have to think about that one.

PG

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

11-07-2014 03:22 PM

I strongly disagree with this statement from loredana.cornea.

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

11-08-2014 11:24 PM

Yes, it's definitely a dataset with a lot of observations!