Wednesday, November 29, 2006

APS 2001 Off-Reserve Data and Chi-Square

Question

I have a question regarding the use of the Chi-square test, this time when using the Aboriginal Peoples Survey.

The researcher notes that "One way to correct the Chi-square for a complex sample is to divide the Chi-square produced, say by SPSS, by the design effect for the sample being used. The result is only approximate, but it works well enough for sorting out what are likely significant relationships."

The answer received for the previous question suggested how to do this for the 2001 Census PUMF of Individuals. Now, he is using the APS 2001 Off-Reserve data set but cannot find a table of design effects, (or conversion factors, or quality factors or anything that looks like values that they can turn into design effect values) in the documentation tha he has for this complex sample. Has he missed something or might there be missing documentation? Do you know where this information might be, or there some other way to do the Chi-square test using the APS?

Answer

I am assuming that the researcher is referring to the PUMF, which does not contain the bootstrap weights and so it is not possible to do this kind of test using that dataset.

My best suggestion would be for the person to get access to the analytical file(s) in the RDC. Failing that, it might be possible to do it through the custom requests route.

No comments:

Post a Comment

Note: Only a member of this blog may post a comment.