Re: st: drawnorm for discrete variables

[Austin Nichols <austinnichols@gmail.com>]

Tyler Frazier <tyfrazier@gmail.com> :
If you have a survey, you should have survey weights.  Why not simply
inflate by the survey weights, or some appropriate multiple of them,
to get the right number of cases, and then -bsample- to get a
synthetic population?  The correlations will be inherited from the
data.  Do you mean to preserve household structure (if your survey
includes multiple people from the same household) or take account of
sample design (e.g. clustering)?

webuse nhanes2, clear
keep houssiz-finalwgt
ta sizplace heartatk [aw=finalwgt], row
g w=round(finalwgt/500)
expand w
ta sizplace heartatk, row
bsample
ta sizplace heartatk, row

On Fri, Apr 30, 2010 at 10:44 AM, Tyler Frazier <tyfrazier@gmail.com> wrote:
> thank you for all your responses and suggestions, now I need to look
> into these possibilities
>
> best regards,
> Ty
>
> On Fri, Apr 30, 2010 at 4:30 PM, Maarten buis <maartenbuis@yahoo.co.uk> wrote:
>> --- On Fri, 30/4/10, Tyler Frazier wrote:
>>> The data will be used in an urban simulation which incorporates a
>>> number of location choice models, regression, and other simple models.
>>>  For now I would like to establish a method for synthetically
>>> generating the data, for use in the larger spatial, framework.  At
>>> this point, I am seeking a plausible solution for running the urban
>>> simulation in order to observe its behavior.
>>
>> I would build such a simulation sequentially, so start by assigning
>> "people" charactericstics that are (largely) exogenous like tribe
>> and sex, and than assign them other characteristics that depend on
>> these, than assign further characteristics based on the prevous ones,
>> etc. For example:
<snip>

*
*   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/