Thanks Steve for your comments. I will be looking into them.
Regards
Larraine
-----Original Message-----
From: [email protected]
[mailto:[email protected]] On Behalf Of Steven
Samuels
Sent: Wednesday, 5 November 2008 3:34 PM
To: [email protected]
Subject: Re: st: Combining household and personal level weights - is
this possible?
Larraine-
I am assuming that you are not doing a multi-level model. If so,
then is your response variable a characteristic of the individual
respondent? of all people in the household? If the first, then use
the person-weight for the respondent. You may still include
household variables as predictors. You need only specify weights
appropriate for unit with the response.
It would be helpful if you would describe concretely the study design
and also some of the regressions you are planning. But here are some
thoughts on possible scenarios:
Example:
1. A single man and a single woman are selected per HH and
interviewed. Your outcome is measured only for the interviewed
person (their age, their health). For this analysis you use the
person-weights, but you can include any HH variables you wish as
predictors.
2. A household respondent is asked questions about the household--for
example what the building material is. In this question, the "house"
is the unit on which the outcome is being measured and the household
weight applies.
3.A household respondent is asked questions about him or herself and
other people in the house--for example, whether each have had a
specific illness. If your outcome includes the other people, then
all , plus the respondent, get the household weight.
I hope this helps.
-Steve
On Nov 4, 2008, at 11:05 PM, Larraine Becker wrote:
> Hi everyone
>
> Another query I have is if you are using survey data with household
> weights as well as person weights, and you want to do a regression
> using
> person and household variables, how do you specify BOTH the weights?
>
> I'm using svrset commands to specify the weights, but as far as I can
> figure out you can only specify one weight variable, and not both.
>
> Any suggestions?
>
> Regards
> Larraine
>
> *
> * 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/
*
* 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/
