4

We conducted 2 similar surveys in parallel to a group of health care facilities. One was sent to managers, one to staff. Some questions were identical, others asked about different areas of the facility. We received responses from groups of facilities that do not completely overlap. So our response rates were the following (approximately).

  • survey a - to staff member at facilities: 70%
  • survey b - to manager at facilities: 81%
  • both surveys: 62%

We are interested in reporting proportions, difference in proportions with (95% CI), and statistical test of difference between proportions for identical question on both surveys (answered by 2 non-overlapping populations)

Martin Bland has a good paper on different approaches...does anyone have SAS code for similar analysis? http://www-users.york.ac.uk/~mb55/overlap.pdf

Jay
  • 41
  • 3
  • Welcome to the site, @Jay. This question seems to be *only* about how to get something done in SAS. As such, it would be off-topic for CV (see our [help page](http://stats.stackexchange.com/help)), but on-topic on [Stack Overflow](http://stackoverflow.com/). If you have a substantive question on the statistical aspects of the different approaches, please edit to clarify; if you just want help w/ the code, flag your Q & we'll migrate it for you (*please don't cross-post, though*). – gung - Reinstate Monica Jul 19 '13 at 02:11
  • I am interested in people's thoughts on different methods, though Blands review suggests his approach is best. Mainly interested in SAS code to execute...how do I flag? – Jay Jul 19 '13 at 02:28
  • Fair enough, could you summarize the different methods you are wondering about, so that people don't have to read the pdf? – gung - Reinstate Monica Jul 19 '13 at 02:29
  • Jay, since you're asking for thoughts on different methods, your post is on topic (i.e. you shouldn't flag it, though for future reference, to flag a post you click 'flag' under the post and explain what you want a moderator to do). A somewhat related post (in that it's about overlapping samples, though not proportions and is not a duplicate; I mention it for some possible relevance) is [this one](http://stats.stackexchange.com/questions/25941/t-test-for-partially-paired-and-partially-unpaired-data) – Glen_b Jul 19 '13 at 02:37
  • Proportions of what? response rates? or a variable in one (or both?) of the surveys? Also, were the target populations identical? – Peter Ellis Jul 19 '13 at 23:26
  • I updated the original post. – Jay Jul 20 '13 at 03:17

1 Answers1

1

Jay,

The following summarises the techniques of Bland, and offers a more modern technique that allows the formation of 95% CI as requested

Test Statistics for Comparing Two proportions with Partially Overlapping Samples http://www.jaqm.ro/issues/volume-10,issue-3/0_BEANDEPA.PHP

Guest
  • 11
  • 2