test for trend , (using survey data)

[Ann Fitzmaurice]

Goo d afternoon

I have data colleced by measureDHS , which is analysed using survey commands
svy linearized, subpop(if year ==1) : tabulate bmi_new3 anc_care2, count row col percent format(%10.0g)

is it possible to do a chi sqaure test for trend, bmi has four ordered categories and anc_care has two categories, skilled and non skilled personnel

i had thought od np trend but not sure this is what i need

can anyone suggest the commands that i need to be able to to this test

i do not think that there is an option within the survey analysis module, so was thinking of a weighted 4 by two , or in some cases a four by three table

thanks

[Rich Goldstein]

one of Stata's FAQs is relevant here: Resources & support >> FAQs >> A comparison of different tests for trend (or, go to state.com, click on support and on FAQs and search for trend)

[Ann Fitzmaurice]

Thanks , i have found the faq, but do not understand how to set this up in stata, i have a four category variable based on the bmi, and a series of othere variables , some with two categories and some with three, i want to see if there is a trend across the bmi, say in antenatal care , but i need to weight the data, i have been using iweight , but cant use this in np trend, is there a way to test fro trend using weighted data

[Rich Goldstein]

it is not clear what your problem is; the FAQ should help you choose what test you want; you then go to the help file for that procedure and see if it accepts weights; nptrend does not; logistic does, including pweights

you also mention that you don't know "how to set this up in Stata"; set what up? if you want logistic regression, just "set it up" as any other logistic regression and add the weight clause; if you want tabulate, again, set it up as any other table and add the weight clause - how is you data currently set up?