Statalist


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: st: PWEIGHT question


From   sjsamuels@gmail.com
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: PWEIGHT question
Date   Fri, 15 May 2009 08:48:19 -0400

Ana is correct. You must use pweights.  If you use -fweights-
(frequency weights) , Stata will think that your sample size is 7
million, not 500.  This, by the way is how SPSS treats weighted data,
and, by doing so,  makes p-values and standard errors unusuable. This
is the cost  of  "simplicity" that Martin wondered about.

A hallmark of weights is that they sum to the population size, so they
should be large numbers, possibly fractional. You can check the
weights by:

egen tot_wt= total(weight_var)
list tot_wt in 1

Who weighted the data? If it is a professional survey organization
they should have documented the sampling design. Th e designmight have
included sampling strata and clusters of telephone exchanges,.  If so,
 use Stata's -svyset- command to indicate the design and use the -svy-
programs to analyze the data.  If there were clusters and you do not
account for them, your standard errors will probably be incorrect.

-Steve

On Fri, May 15, 2009 at 4:14 AM, Ana Gabriela Guerrero Serdan
<ag_guerreroserdan@yahoo.com> wrote:
> You probably need to use pweight, here is a link that explains the difference between fw, pw and aw:
>
> http://www.cpc.unc.edu/services/computer/presentations/statatutorial/example30.html
>
>> regards,
> Gaby
>
>
>
> --- On Fri, 5/15/09, Charles Man@CCR <ccrman@cityu.edu.hk> wrote:
>
>> From: Charles Man@CCR <ccrman@cityu.edu.hk>
>> Subject: st: PWEIGHT question
>> To: statalist@hsphsun2.harvard.edu
>> Date: Friday, May 15, 2009, 1:50 AM
>>> I have a small telephone survey dataset of 500 cases which
>> were randomly drawn
>> from a population of 7 million people. The dataset was
>> weighted according to
>> income and sex distributions of the population. In Stata,
>> do I need to include
>> [pw=weight] in all analyses? In SPSS, people simply choose
>> WEIGHT BY and don’t
>> have to be concerned about the weight type. However, in
>> Stata, we have to
>> determine which weight type is appropriate. Is it Pweight
>> or Fweight useful in
>> my situation? Which one is the most appropriate?
>>

*
*   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/



© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index