complex sample

[Dominic Lusinchi]

Dear colleagues,



I am working (or should I say attempting to) on a data set from the National
Comorbidity Survey.



The data set comes with weights. However, the data set provided by the
client does not include the variables to take into account the complex
sampling that was used to obtain the data. Without these variables the
standard errors will be (most likely) underestimated thus increasing the
risk of a Type I error.



Obtaining these variables should not be too much of a problem, since the
FAQs page for the survey (http://www.hcp.med.harvard.edu/ncs/faqncs.php)
states that "We also added variables STR (1-42 sampling strata) and SECU
(1-2 random halves of each stratum) for purposes of computing design-based
standard errors. (.) However, for users who are familiar with resampling
methods, these two variables are all you need."



My issue is this: can I perform the analysis using the variables just
mentioned in SPSS, or can it be done only using the Complex Samples module
(which I do not have)? I know that bootstrapping can be performed using OMS
but is it applicable to my case here?



I should add two things: 1) I am using version 13 and 2) I am performing a
binary logistic regression.



Any advice would be much appreciated.



Thanks in advance,

Dominic



**********************************************

Dominic Lusinchi

Consultant



Far West Research

Statistical Consulting



San Francisco, California



Telephone: [phone]

Fax: [phone]

Email:  <mailto:[hidden email]> [hidden email]

Web:  <http://www.farwestresearch.com> www.farwestresearch.com

**********************************************