Re: st: Question on interpretation of survey data: from the mean to the total

[Stas Kolenikov <skolenik@gmail.com>]

I would go with

svy : mean cars

and

svy : total cars if (household_head)

if "cars" is the variable where the number of cars in the HH is
stored. It does not matter what was the purpose of the data
collection, but you need to have a clear notion of the units of
sampling, weighting, and analysis. If the latter matches the ave #
cars times expanded # of HHs, that's great, but you won't be able to
get the standard errors around it. With the above syntax, you will.

On Sun, Sep 12, 2010 at 8:09 PM, Laurie Molina <molinalaurie@gmail.com> wrote:
> In my dataset each observation is a household and the weights are for
> households.
> In the survey there is a question about the number of cars of the household.
> So i am interested in the average number of cars of the HH.
> An also i am interested in the total number of cars owned by HH.
>
> To my understand it is ok to calculate the average number of cars of
> the HH, even with
> summarize cars[weight=factor].
>
> But if i get that mean, and i multiply it by the (expanded) number of
> observations (HH), is it ok to say that it is the total number of cars
> of the HH in the survey?
>
> Thank you very much again!
>
>
> On Sun, Sep 12, 2010 at 7:13 PM, Steve Samuels <sjsamuels@gmail.com> wrote:
>> --
>>
>>  In advance I would say that you should create a data set in which
>> each observation is a HH and the weight is an appropriate HH weight.
>> However to advise you further with some assurance, we  would need to
>> know how the weights were computed-was there post-stratification or
>> "raking", for example and details about the design (clusters, stages
>> is too vague). including the -svyset- statement.  Your current
>> approach is almost certainly wrong, because the denominator for -svy:
>> mean- will be individuals, not HH; it will be biased, because larger
>> HH will contribute more observations.
>>
>> Steve
>>
>> Steven J. Samuels
>> sjsamuels@gmail.com
>> 18 Cantine's Island
>> Saugerties NY 12477
>> USA
>> Voice: 845-246-0774
>> Fax:    206-202-4783
>>
>>
>>
>> On Sun, Sep 12, 2010 at 2:50 PM, Laurie Molina <molinalaurie@gmail.com> wrote:
>>> Dear all,
>>> I have a question on the interpretation of survey data.
>>> There is a survey in which the analysis units are the households and
>>> its members.
>>> Using the command svyset i have set the design characteristics of the
>>> survey: clusters, two stages, stratification and probabilistic.
>>> And i have asked stata for the mean of the number of cars in the households.
>>> Is it all right if i multiply the mean of the number of cars in the
>>> households by the expanded number of households to get the total
>>> number of cars of the households?
>>> Or does it depend on wheter the survey was designed to get the total
>>> number of cars of the households?
>>> Thank you very much in advance.
>>> Laurie
>>> *
>>> *   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/
>



-- 
Stas Kolenikov, also found at http://stas.kolenikov.name
Small print: I use this email account for mailing lists only.

*
*   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/