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
- /
- Proc PSMatch: Error: Floating Point Zero Divide. &...

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

06-08-2017 03:23 PM - edited 06-09-2017 09:44 AM

I am getting the following error/warning when Proc PSMatch used on SAS9.4 TS Level 1M4, 64bits Windows 7Pro Platform. Basic model summary is:

- Used Greedy K=2 match with caliper = .10
- LPS (logit of PS) used with 'no exact' option and 'no weight' option
- Independent variables are either categorical or nominal
- No continuous variable used
- Event/Cases: N= ~42,000

**Warning from SAS Log:**

Spoiler

WARNING: The maximum likelihood estimates for the logistic regression model might not exist. The

maximum likelihood estimates are based on the last maximum likelihood iteration.

maximum likelihood estimates are based on the last maximum likelihood iteration.

Spoiler

ERROR: Floating Point Zero Divide

*ERROR: Termination due to Floating Point Exception*

**Total N** = 1,289,000 (rounded).

Looks like main issue is 'Number of Observations' used in a model. As I tested with less observations, it worked. Details below:

*----------------------------------------------------------------------------------------------------;

Same code works without any issues or warning when Total N = 915,000

*----------------------------------------------------------------------------------------------------;

When sample size incrased to, i.e. N=1,040,000 with same set of codes, the following warning sign appears but still produces result:

**WARNING**: The maximum likelihood estimates for the logistic regression model might not exist. The

maximum likelihood estimates are based on the last maximum likelihood iteration.

*----------------------------------------------------------------------------------------------------;

Again increased sample size to: N=1,164,000, it is still working with warning sign.

WARNING: The maximum likelihood estimates for the logistic regression model might not exist. The

maximum likelihood estimates are based on the last maximum likelihood iteration.

Again increased sample size to: N=1,286,000 it is still working with warning sign.

WARNING: The maximum likelihood estimates for the logistic regression model might not exist. The

maximum likelihood estimates are based on the last maximum likelihood iteration.

The magic number for not working the procedure is somewhat close to: *1,289,000 (rounded) but greater than 1,286,000.*

Not sure what to make of it. Any suggestions appriciated!

Thx,

APU_007