topic Re: Number of people in a regression model in Statistical Procedures

Number of people in a regression model

CEdward — Thu, 10 Dec 2020 04:17:54 GMT

Good day,

I saw this post previously and am wondering if the opposite is done: https://communities.sas.com/t5/Statistical-Procedures/Identifying-Unused-Observations-in-PROC-PHREG/td-p/143602.

Suppose I have a panel dataset, how can I determine how many people were included in my model?

Re: Number of people in a regression model

PGStats — Thu, 10 Dec 2020 04:36:41 GMT

Same logic:

Use the fact that residuals cannot be calculated for unused observations. Try adding the statement

output out=resOut resmart=resmart;

to the phreg procedure, and then

proc print data=resOut; where resmart is NOT missing; run;

to print the obs used, or

proc sql;

select count(resmart) as numberUsedInModel

from resOut;

quit;

to get the number of obs used.

Re: Number of people in a regression model

CEdward — Thu, 10 Dec 2020 04:42:28 GMT

PGStats

Thank you.

I want to get the number of distinct individuals used rather than observations. I have used that method for observations already.

Re: Number of people in a regression model

ballardw — Thu, 10 Dec 2020 05:04:25 GMT

@CEdward wrote:

PGStats

Thank you.

I want to get the number of distinct individuals used rather than observations. I have used that method for observations already.

Does your data set have an individual identifier? Or group of variables that identify an individual? Then something like this gets the individuals, so you could use that as the basic for counting as a subquery.

Proc sql;

select distinct <variable(s) to identify an individual>

from residualdataset

where not missing (residualvariable)

;

quit;

Re: Number of people in a regression model

CEdward — Thu, 10 Dec 2020 05:06:58 GMT

I suspected that was the solution. Thank you.
As an added safety check, I noticed a post in that original link which says that using the fact that residuals are not always calculated for lines with missing is not bullet proof. Would someone be able to confirm that?

Re: Number of people in a regression model

ballardw — Thu, 10 Dec 2020 05:13:12 GMT

In that post @JacobSimonsen posted an example for one specific regression with a specific data set.

Other regressions might have a similar behavior depending on data, regression procedure and options used.