David Greenberg <[email protected]>:
I already made my own view clear at
http://www.stata.com/statalist/archive/2008-01/msg00472.html
but I can't think of a model that I would use where the notion of a
superpopulation is not necessary, much less not "ridiculous" (your
word, not mine). Suppose you have data on every kid in the school, not
a survey, and every school in the district, and you want to test for
some form of sex discrimination in assignment to a program. Well,
just see if more girls than boys (or vice versa) are assigned, and
there is your evidence of discrimination, right?
On Jan 18, 2008 2:49 PM, David Greenberg <[email protected]> wrote:
> There is an old debate going back to the 1970s about the meaningfulness of statistics when dealing with entire populations. If I recall correctly, Judith Tanner edited a book of papers on the subject. Proponents of testing in this circumstance say that we can think of the population of countries as having been sampled from an imaginary larger population of all possible countries, but I think this is ridiculous. We know that the existing set of countries was generated by historical processes (conquest, secessions, and the like) that wasn't random. With time series data it may make sense to imagine a hypothetical random generating process from which a certain stretch of time has been sampled. Ther may also be circumstances where something like bootstrap standard errors could be informative. Suppose you are studying all the children in a school. You would not have a simple random sample, but might still want to know how sensitive your results are to the possibility that a few
ch
>
> ildren were not there on the day you passed out your survey instrument because they were sick or truant. David Greenberg, Sociology Department, New York University
>
*
* 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/
