Re: st: Multistage sampling svyset

[Steve Samuels <sjsamuels@gmail.com>]

Thanks, James.   Occasionally people use the words "randomly selected"
to describe something that was more involved, and I wanted to make
sure that the design was correctly specified.

Steve

On Thu, Jun 10, 2010 at 7:03 PM,  <James.Valcour@med.mun.ca> wrote:
> Hi Steve,
>
> Thanks for the explanation on the addition of HH addition to the variance and why it's not needed.
>
> We had a complete sampling frame of all the houses in the communities.  These are very small rural communities with roughly anywhere from 20-500 households.  Households were randomly selected from the frame for each community with number of households sampled in each community proportionate to it's size.  Once households were selected a member of the household who was eligible for the study (i.e. older than 16 years of age) was randomly selected from within the household.  We had complete information about all the members in each household.
>
> James
>
> --
> Dr. James Valcour
> Division of Community Health and Humanities
> Faculty of Medicine
> The Health Sciences Centre
> Memorial University of Newfoundland
> St. John's, NL, Canada
> A1B 3V6
> Tel:  709-777-2237
> Fax: 709-777-7382
> Email: james.valcour@med.mun.ca
>
>
>
> ________________________________
>
> From: owner-statalist@hsphsun2.harvard.edu on behalf of Steve Samuels
> Sent: Thu 10/06/2010 6:01 PM
> To: statalist@hsphsun2.harvard.edu
> Subject: Re: st: Multistage sampling svyset
>
>
>
> You still haven't answered my question about the design, James:
> exactly how were HH selected?
>
> To answer your question: the sampling fraction within HH is not
> ignored; it should be reflected in the sampling weight. (Check to be
> sure that it is.)   The -fpc- applies only to sampling units for which
>  a contribution to variance can be estimated.  As you have only one
> observation per family, there is no within-family variance  component
> to estimate,  hence no fpc.  In fact  even if several individuals had
> been selected within each hh, fpcs would  usually not be computed,
> because they make no practical difference in the standard errors and
> because they would be invalid for  many sub-population estimates.
>
> Steve
>
> On Thu, Jun 10, 2010 at 3:58 PM,  <James.Valcour@med.mun.ca> wrote:
>> Thanks for the more detailed read of my original message ;)
>>
>> I tried as you suggested, and it works (thank you!).  Does that not ignore the selection probability of the individuals within the household?
>>
>> I also tried Michael's suggestion of using the singleunit option.  I get the same standard error estimates using that option and the more complex original syntax.
>>
>> Regardless, I am now getting the estimates I've been looking for.
>>
>> Thank you to you both.
>>
>> Cheers,
>> James
>>
>> -----Original Message-----
>> From: owner-statalist@hsphsun2.harvard.edu [mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of Steve Samuels
>> Sent: June 10, 2010 4:48 PM
>> To: statalist@hsphsun2.harvard.edu
>> Subject: Re: st: Multistage sampling svyset
>>
>> OK, I read your email in detail.  Let pcomm = (N hh in the stratum)/n
>> hh in the sample from the stratum)
>>
>> _If_ the survey was done as you describe the -svyset- should go:
>>
>> svyset houseid [pw=pwgt], strata(pi4) fpc(pcomm)
>>
>> The proportion of people selected in the HH (the last stage)  affects
>> only the sampling weight.
>>
>> Please verify how the HH were selected  Was there a list of HH from
>> which HH were sampled?   Or, was there an intermediate sample of areas
>> (such as blocks) drawn first, with HH sampled only in the selected
>> areas?
>>
>> Steve
>>
>> On Thu, Jun 10, 2010 at 3:04 PM, Steve Samuels <sjsamuels@gmail.com> wrote:
>>> James, what is "pi4"? What is the variable which identifies the
>>> community?  How did you compute the sampling weight "pwgt"?  (show an
>>> example, as well as code.)
>>>
>>> Steve
>>>
>>>
>>> On Thu, Jun 10, 2010 at 2:33 PM,  <James.Valcour@med.mun.ca> wrote:
>>>> I'm having a problem using the survey commands to analyzing a survey
>>>> that used a complex sampling design.
>>>>
>>>> Eleven communities were sampled (these were all the communities in the
>>>> study area).  Within each community households were randomly selected
>>>> from all available households.  Within each household one person was
>>>> randomly select from all members of the household that were eligible to
>>>> be in the survey (we had a list of everybody eligible in the house).
>>>>
>>>> This is the code I used to set the data:
>>>>
>>>> svyset houseid [pw=pwgt], strata(pi4) fpc(p_household) || _n,
>>>> fpc(p_person)
>>>>
>>>> where
>>>> houseid is an identifier for the house in the community; a mix of the
>>>> community id number and the house number
>>>> pi4 is an identified for the community (1-11)
>>>> p_household is the probability of selection of the household in the
>>>> community
>>>> p_person is the probability of selection of the person in the household
>>>>
>>>> The problem I'm getting is when I'm trying to estimate the proportion of
>>>> study with certain ailments.  I'm not getting confidence intervals
>>>> included in the estimates.  The message that is given is " Note: missing
>>>> standard errors because of stratum with single sampling unit."
>>>>
>>>> I've tried other ways of specify the svyset command and tried
>>>> simplifying to the point of just assuming that household were sampled
>>>> but I get the same error.  (for the record here's that command "svyset
>>>> _n [pw=pwgt], strata(pi4) fpc(p_household)")
>>>>
>>>> Any help would be appreciated,
>>>>
>>>> Thanks!
>>>> James Valcour
>>>>
>>>> --
>>>> Dr. James Valcour
>>>> Division of Community Health and Humanities
>>>> Faculty of Medicine
>>>> The Health Sciences Centre
>>>> Memorial University of Newfoundland
>>>> St. John's, NL, Canada
>>>> A1B 3V6
>>>> Tel:  709-777-2237
>>>> Fax: 709-777-7382
>>>> Email: james.valcour@med.mun.ca
>>>>
>>>>
>>>>
>>>> *
>>>> *   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/
>>>>
>>>
>>>
>>>
>>> --
>>> Steven Samuels
>>> sjsamuels@gmail.com
>>> 18 Cantine's Island
>>> Saugerties NY 12477
>>> USA
>>> Voice: 845-246-0774
>>> Fax:    206-202-4783
>>>
>>
>>
>>
>> --
>> Steven Samuels
>> sjsamuels@gmail.com
>> 18 Cantine's Island
>> Saugerties NY 12477
>> USA
>> Voice: 845-246-0774
>> Fax:    206-202-4783
>>
>> *
>> *   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/
>>
>
>
>
> --
> Steven Samuels
> sjsamuels@gmail.com
> 18 Cantine's Island
> Saugerties NY 12477
> USA
> Voice: 845-246-0774
> Fax:    206-202-4783
>
> *
> *   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/
>
>
>



-- 
Steven Samuels
sjsamuels@gmail.com
18 Cantine's Island
Saugerties NY 12477
USA
Voice: 845-246-0774
Fax:    206-202-4783

*
*   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/