Re: st: RE: RE: Applying weights to Survey Data

[Steven Samuels <sjsamuels@gmail.com>]

ajjee-

You should use the probability weight [pweight=] option. The sample is  
likely to have had strata and multiple stages of sampling, with the  
first stage often called the "primary sampling unit".  If so, you will  
get correct standard errors only if you use Stata's survey programs,  
starting with -svyset-  and, ending, perhaps, with -svy logistic-.   
You should  understand the the survey design before trying to write  
the -svyset- command; so you must also read the study documentation.

Steve

On Nov 26, 2010, at 6:17 AM, Nick Cox wrote:

There might be enough information for a -svy-savvy person (not me) to  
suggest syntax for you, but either way my advice remains to read the - 
svy- documentation.

Nick
n.j.cox@durham.ac.uk

ajjee

Thanks Nick about your advice to use the word "a dataset".

The description of weights is as follows.

Description of weight: This POST-STRATIFICATION WEIGHT is based on a
comparison for each sample with the respective universe description.  
As such
in all countries, gender, age, region and size of locality are  
introduced in
the iteration procedure carried out by the fieldwork institutes.

Nick Cox wrote:

> You don't say how these weights are calculated, so it is difficult to
> advise. But as the documentation does tell you, frequency weights  
> must be
> positive integers, so they are certainly not fweights.
>
> You should check out the -svy- documentation.
>
> By the way, although perhaps just used informally, the wording "a  
> data"
> meaning "a dataset" wouldn't get past me if I read it in a draft  
> presented
> to me as teacher, supervisor, reviewer or editor. I don't regard it as
> acceptable technical English.

ajjee

> I am working on a survey data in which weights are given for each
> respondent. Now I am confused  how to apply thses weights in the - 
> logit-.
> My
> data looks like:
>
> cnty	      id	    wt	             trd	    edu	 sex	      age	
> inc
> FRA	   6407	   0.7749	       F	     16	MALE	        18	      .a
> FRA	  7026	   0.0964	       F	     18	MALE	        34	       +
> FRA	  6309	   0.5040	       F	     17	FEMALE       25	      +
> FRA	  5430	   0.5331	       A	     17	FEMALE	 33	       -
> FRA	  5106           0.9501	      .b	     14	FEMALE	 83       	-
> FRA	  6717	   1.1677	       F	     18	MALE  	 48	       +
> FRA	  6813	   0.8413	      .b	     20	FEMALE	 35	        -
> FRA	  5185	   1.3552	       F	     18	MALE	         43     	+
> FRA	  6045	   1.0003	       F	     23	MALE  	 23      	-
> FRA	  5954	   2.0298	       F	     14	FEMALE	 59     	+
> FRA	  6754	   0.8034	       A	     17	MALE	          31    	.a
> FRA	  5748	   0.8631	       F	     23	MALE   	40     	-
> FRA	  6209	   0.9786	       A	     17	MALE  	30     	+
> FRA	  6120	  1.1140	       F	     20     FEMALE	        23	        -
> GBR	  6181	   0.7317	       A	     23	MALE	        40     	+
> GBR	  6267         0.7213	       F	     25	MALE	         23    	.a
> GBR	  5223	  0.9928	       .b	     20	MALE	         31	       -
> GBR	  6670	  1.1046	       A	     22	FEMALE	25	           +
> GBR	  5435	  1.4809	       F	     25	FEMALE	78	        +
> GBR	  5931        1.4809	      .b	     16	FEMALE	63	          .a
> GBR	 5219	         0.4395	       F	     18	FEMALE	53	          +
> GBR	  5130	1.0084	       F	     23	FEMALE	50	           -
>
> and my regression is
>
> logit trd edu sex age inc,cl(cnty). Should I use (fweigt=wt) or
> (aweight=wt)

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/