Chi-square test in bivariate analysis

[Pintu Paul]

Hi,
I am confused using sample weight in bivariate analysis. Is it we need to use sample weight in chi-square test? or we just can test without using sample weight? I am using 'tabout' command in STATA and when I am using weight in chi-square test the value of chi2 is goes down and the value of chi2 is pretty high without weight. I am asking which one is appropriate when we do chi2 test in bivariate analysis?
Thank You!

[Steve Samuels]

I'm afraid that neither answer will give you a correct analysis. You apparently have sample survey data. Analyses of survey data must take into account not only that there are sample weights, but that sampling may be stratified, with on or more stages of clustered sampling.

Stata has a complete suite of commands to set up analyze survey data. You: 1) Use the command svyset to specify the survey characteristics to Stata; then 2) use a "survey-aware" command, one which uses the svy prefix. In your case, it is likely to be svy tabulate.

If you are unclear about sampling concepts, there are many good texts and online descriptions. Some links:

* http://www.statisticalassociates.com/sampling.pdf

* http://www.statcan.gc.ca/edu/power-p...214895-eng.htm

* http://www.restore.ac.uk/PEAS/theory.php.

Study one or more of the references above and at the Remarks and Examples in the Manual entry for svyset Then describe for us the design of your smple.

Important: before you post, read the Statalist FAQ https://www.statalist.org/forums/help, especially #12 . Follow its directions for how to ask good questions. Always show commands and results, in between CODE delimiters, as described in FAQ 12

[Marcos Almeida]

Steve gave excellent advice.

Just hazarding a guess, maybe you mean chi-square with the whole data versus aggegated date.

If you have aggregate data, either you use frequency weight in the command, or you - expand - the data before performing the chi-square command.

Hopefully that helps.