on 21/11/2002 2:46 pm, ann e fitzmaurice at [email protected] wrote:
> i have the following output from the survey commands procesure
> The psu is a self computed variable based on the number of times
> the woman appears in the data set
>
> my question relates to the p value
> . svyset psu sib_no1
>
> . svytab sib_grp v106a if level1 ==1, pearson row count format (%9.2f)
>
> pweight: weight Number of obs = 12531
> Strata: <one> Number of strata = 1
> PSU: sib_no1 Number of PSUs = 9
> Population size = 12489.897
The PSU looks wrong. Cannot have a dataset of 12531 observations that has
just 9 PSUs. The PSU identifies the primary sampling unit - the mother, I
suspect. Each mother should have an ID which appears on the person record to
identify which children came from which mothers.
Your approach to the table also looks wrong. You have one ordinal variable
(extent of education) and possibly a second. A Chi-squared test does not
take this ordering into account. If your predicted variable is ordinal,
check out -svyologit-. The other possibility is that it represents a
baseline category and two outcomes of interest. In which case you could look
at -svyologit-. The Chi squared doesn't really test any interesting
hypothesis. 'Significant' or not, there is an element of 'so what?' in its
interpretation.
Ronan M Conroy ([email protected])
Lecturer in Biostatistics
Royal College of Surgeons
Dublin 2, Ireland
+353 1 402 2431 (fax 2329)
--------------------
*
* 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/
