Re: st: sample survey analysis without PSU info

[Steven Samuels <ssamuels@albany.edu>]

Carlos, with only stratum information and weights, you cannot  
estimate standard errors. Correct standard errors depend mostly on  
the number of PSU's and on the variation between them.  Any standard  
errors that ignore the design will be completely useless. Almost  
always they will be lower than the real standard errors and they  
could be much lower.

However many surveys that do not identify PSU's supply other means of  
computing standard errors.  For example, they may identify replicates  
to use with replicated-based variance methods.  Or, the documentation  
may contain formulas or design effects for approximating the standard  
errors.  Without such additional information, you have _no_ remedy.


A good reference is Sharon Lohr's book  Sampling: Design and  
Analysis, Duxbury, 1999.

Steve

If not, there is no remedy.
On Mar 12, 2009, at 11:42 AM, Carlos Dias wrote:

> Dear Statalisters,
> I am preparing to analyse a large sample survey data file (about 40000
> individuals where all individuals in families are surveyed). This
> comes from a multistage, cluster sample survey where strata=regions,
> PSU=counties, and sample weights calculated using Horwitz-Thompson
> methods in R and then post-stratification with marginal adjustements
> for age and sex.
>
> I have sample weights and strata information for each individual but
> no information on the PSU (counties within strata) of each individual.
>
> Could someone direct me to references on the risks and ways to
> minimize them when I conduct the analysis without cluster information
> on the svyset command?
>
> Thank in advance
>
> Carlos Dias
> National School of Public Health, New University of Lisbon
> Lisbon, Portugal
> *
> *   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/