proc surveyfreq question

[Addison]

Hello, 

I have a question about using proc surveyfreq procedure to examin the relationship between two categorical variables.

Currently I use chisq to perform chi-square statics but my colleague use chisq1 to performa chi-square statics.

The same dataset and same variables, when I run the proc surveyfreq procefure with chi-sq and obtained p value 0.0503. My colleague will get a different p value (0.0463) by using chisq1 to run chi-square statistics.

I know chisq1 is requested Rao-Scott modified chi-square test. My question is how do I know which chi-square statictics I should apply?

Does anyone know what's difference between chisq and chisq1? 

[Rick_SAS]

I think the PROC SURVEYFREQ doc provides these answers if you know where to look.:

1) Go to the SAS documentation

2) Click the procedure you want information about. In this case, PROC SURVEYFREQ.

3) Look under the "details" tab. This is where statistical formulas are usually presented and discussed.  In this case, select "Statistical Computations."

4. You will see many sections that explain how the SURVEYFREQ procedure works. Select "Rao-Scott Chi-Square Test."

The doc does not include the string 'chisq1'.  However, it does discuss the first-order Rao-Scott chi-square statistics for the goodness-of-fit test.  It includes links to the syntax section (how do you get the test?), links to the literature (who invented the test? When?) and links to other sections of the doc.

The doc tells me that CHISQ1 option is very old and the most recent version of SAS no longer uses that name. The option is now called the CHISQ(MODIFIED) option. The syntax section for the CHISQ option describes the difference between the first-order Rao-Scott test and the modified test.

[ballardw]

From the online documentation:

Two forms of the design correction are available for the Rao-Scott tests. One form of the design correction uses the proportion estimates, and you request the corresponding Rao-Scott chi-square test with the CHISQ option. The other form of the design correction uses the null hypothesis proportions. You request this test, called the Rao-Scott modified chi-square test, with the CHISQ1 option.

There is much more detail in the online help if you want to work through the algebra.