# Re: st: Determining Sample Size

 From Jeph Herrin To statalist@hsphsun2.harvard.edu Subject Re: st: Determining Sample Size Date Thu, 02 Nov 2006 08:25:32 -0500

```Perhaps if you can identify 1) your outcome measure and 2) your
research question? Without these it is difficult to offer any
suggestions on how to proceed. For instance, without knowing
your outcome measure, it is not possible (at least, not meaningful)
to advise on whether you should be comparing proportions or means.

Jeph

Melissa Cortina wrote:
```
```I have a population of 5,000 checks broken into 5 categories based on
dollar value. Each stratum is 1000 checks. I need to select a sample of
checks from each stratum to actually research. Can I get Stata to tell
me the minimum sample size from each stratum to give me a 90% confidence
level?  I've been trying using the "sampsi" command. The biggest issue I
have is that I know the actual population and Stata doesn't seem to take
that into consideration. Is there a better command I could be using?

If not, I have a couple of questions about the use of the "sampsi"
command. I know that I have a one-sample test. I'm not sure if I would
want to test for mean or equity of proportions.

I feel that I have more information to do a test of mean. I can use a
null hypothesis of the actual population mean. I would use a two-sided
test with alpha = .05 and a power of 90%. I can use the population SD as
well. Would I use the population mean again as the expected mean in the
sample?

"Sampsi 224.6 2935.516, sd(16979.5) onesam" Returns "416" Which seems
extremely large. That is using the stratum mean (224.6) and the
population mean (2935.516) and population sd(16979.5)

I could also do a test of proportion with my null hypothesis as 70% of
the checks. Could I use an expected proportion (from a past audit,
discretion, etc) as my proposed proportion?

My goal is to end up with a sample that is as close to my real
population as possible. (As I suppose is everyone's.) I'm not entirely
sure that the mean is entirely important to me, because the dollar
amounts of the checks are not important once they are within strata. I
think that a test of proportion may be better because I can use a higher
expected proportion in a lower dollar amount stratum than in a higher
amount one. Basically I want to be able to show statistically that the
sample size number we have chosen is within a 90% confidence of
capturing information from the whole population.

Thank you.

~Melissa

"This message contains confidential information. Unless you are the addressee (or authorized to receive for the addressee), you may not copy, use or distribute this information. If you have received this message in error, please notify us by replying to the message and deleting it from your computer.  Thank You.   Kelmar Associates, LLC."

*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/

```
```*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/
```