proc means - confidence interval - which tables are used?

[xxformat_com]

Hi,

When computing confidence interval (CLM), proc means seems to use

- the Student table for small values of n (eg. n=44 > 2,0167 for a 95%CI)

- the Normal table for large values of n (eg. n=4400 > 1,96 for a 95%CI)

Do you have references in the SAS online doc which would indicate which table is used depending on which value of n for CLM, LCLM and UCLM?

Best Regards,

[WarrenKuhfeld]

p values are not looked up in tables. SAS has software to compute p values for a bunch of distributions given a particular set of parameters. Normal and t are not much different for large n.

[xxformat_com]

Thanks. But I'm looking for the computation method used by the system.

Of course, table is just the manual way of accessing to the information. I guess I should have make it clearer using distribution used.

But my point is: which one is selected based on which criteria by the system and where it is documented.

[PaigeMiller]

---

@xxformat_com wrote:

When computing confidence interval (CLM), proc means seems to use

- the Student table for small values of n (eg. n=44 > 2,0167 for a 95%CI)

- the Normal table for large values of n (eg. n=4400 > 1,96 for a 95%CI)

---

The t-table value for 4399 df is, to 2 decimal places, the exact same value as the normal distribution.

[xxformat_com]

Hi,

Thanks but my question is not to compare the student distribution with the normal distribution for large df. My question is where is specified in the SAS online documentation which tables are programmed internally in the system.

[WarrenKuhfeld]

As I said, there are no internal tables to document.  Mathematical programming statements generate the results.

[xxformat_com]

As as said, my point is not whether it is a table or a formula. My point is to find the definition used internally.

[PGStats · Accepted solution]

Here:

https://documentation.sas.com/?docsetId=proc&docsetTarget=p0v0y1on1hbxukn0zqgsp5ky8hc0.htm&docsetVer...

[xxformat_com]

Thank you. That's exactly what I was looking for 🙂