Calculating sample - Two stage stratified cluster sample

[Ulysse Sante]

Hello Stata users, I am wondering if you can help me. We want to carry out a survey to assess perceptions of quality and access to medical education of last year students.

Based on our research, we have decided that a two stage stratified cluster sample suits our needs in terms of costs and logistics.

We have 37 medical institutions in the country so the first stage will be stratified and will have 4 strata
Strata 1 - Public and Urban
Strata 2 - Public and Rural
Strata 3 - Private and Urban
Strata 4 - Private and Rural

The institutions in each strata will be selected randomly with the number of institutions to be selected from each stratum proportional to the total stratum size.

For the second stage we will randomly select a number of students from registry lists.

Can someone please shed light on how to calculate the sample size (number of PSUs and SSUs?

Thanks
us

[Steve Samuels]

The FAQ ask that you link to cross-posted messages. Please do so, on both ends.

[Ulysse Sante]

http://stats.stackexchange.com/quest...cluster-sample

[Steve Samuels]

I forgot that I'd answered a similar question on Stack Exchange: http://stats.stackexchange.com/questions/122274/numbre-of-cluster-in-pps-sampling-method/123961#123961. If schools vary greatly in size, then sampling proportional to size is probably better than your current plan to allocate to strata proportional to the number of schools in each stratum. In fact, with such a small population of schools, this could be difficult to do. If so, I would list schools in stratum order and use the systematic sampling method of Madow, described in Cochran (1977, p. 265) and implemented in Stephen Jenkins's samplepps (SSC). See page 268 of Kish (1965) "Specific Cost Function for Cluster Samples". See also Lohr, 2009, Section 5.3


As that post makes clear, you require some estimates of relatives costs and of between- and within- cluster variation. If information is not available from similar studies, then you will need to guess or to do a pilot test to estimate them. A pilot study is a very good idea in any case, as it will identify unexpected problems. Note that "cost" might be in monetary units or in man-power units (person- hours). Often it will include both, with person-hours translated to monetary units.

These costs will depend on mode of contact. In my experience, surveys of students can have low response rates. Pretesting and pilot testing of every study component is needed to refine the contact method, to ensure that the questionnaire will be clearly understood and easy to answer. In general, it is better to have a smaller sample size, with high response, than a larger sample and poor response. In fact, you should consider ways of contacting a subsample of non-responders.

You may be fortunate enough enough to have a well-tested and validated questionnaire. If not, I will just say that writing good survey questions is very hard, so I hope that your colleagues include someone with experience in this area.

For good advice on non-response and questionnaire issues, see Groves et al. (2009).

References:

Cochran, William G 1977 Sampling techniques, Wiley, NY

Groves, Robert M., Floyd J. Fowler, Mick P. Couper, James M. Lepkowski, Eleanor Singer, and Roger Tourangeau. 2009. Survey methodology, Second Edition. Hoboken, N.J.: Wiley.

Kish, Leslie. 1965. Survey sampling. New York: Wiley. Lohr, Sharon L. 2009. Sampling: Design and Analysis. Boston, MA: Cengage Brooks/Cole, Section 5.3
Groves, Robert M., Floyd J. Fowler, Mick P. Couper, James M. Lepkowski, Eleanor Singer, and Roger Tourangeau. 2009. Survey methodology, Second Edition. Hoboken, N.J.: Wiley.