weighted survey data analysis commands: pweight and aweight?

[suveyda karakaya]

Dear All,

Since this is first time I am doing survey analysis with weighted data, I am not sure whether I run the logit regs properly and some commands in stata dont work with svy syntax. For ex, I can't do factor analysis with pweight option, therefore I used aweight option. I have some questions:

Is there any difference between aweight and pweight? Is it ok to use aweight command when pweight does not work?

Since I use svyset command with lineralized option I dont have to use robust cluster option for std errors right?

Are these commands to run logit regression correct:
svyset mah [pweight=w_yerlesim], strata(bolge) vce(linearized) singleunit(missing)

pweight: w_yerlesim
VCE: linearized
Single unit: missing
Strata 1: bolge
SU 1: mah
FPC 1: <zero>

svy linearized : logit nonviolentresistance male age politicaltrust orgmember2 polinterest Education Income

Many thanks!

[Marcos Almeida]

Hello Suveyda,

Welcome to the Stata Forum. You are supposed to apply proportional weights under a survey design. Please use the CODE delimiters to post the commands in Stata. That said, your first command seems to me quite correct. Just a slight comment: since the options "vce(linearized)" and "singleunit(missing)" are both the default options, you don't even need to type them. Also, I suggest you exclude "linearized" from the "svy" command.

Finally, I recommend you to take a close look at the Stata manual. This is the best way to get to grips with the matter and get used with the mainstream commands.

Hopefully that helps.

Best,
Marcos

[Steve Samuels]

I think that Marcos intended

You are supposed to apply probability weights under a survey design.

[Marcos Almeida]

Steve Samuels : You're absolutely right, Steve. I really meant probabiblity weights. I guess my mind is playing tricks on me. Thank you for the correction. Best, Marcos.

[suveyda karakaya]

Hello Again,

Thanks a lot. When I run the logit reg, populations size is decimal and greater than number of obs. Is it normal or am I doing a mistake? I copied the stata output below. Am I running the regs correctly or not???

[Marcos Almeida]

Hello Suveyda,

Concerning your query in #5 (over the estimated size of the population): that is part and parcel of the survey design.

You may this and all background information here:http://www.stata.com/manuals13/svy.pdf

Best,

Marcos

[Richard Williams]

Here is a brief Intro to SVY handout I just went over in my Grad Stats class that might be helpful.

http://www3.nd.edu/~rwilliam/xsoc73994/SvyCautionsX.pdf