Population Assessment of Tobacco and Health

Using longitudinal weights for data in long format Tips for Using the PATH Study Data User Forum

Return to the Population Assessment of Tobacco and Health (PATH) Study Series page.

6 posts / 0 new
Last post
albert.ksinan
Using longitudinal weights for data in long format

Hello,

our analysis uses PATH data waves 1-4 with appropriate longitudinal weights. Upon reading the manual, it was my understanding that the VARPSU and VARSTRAT were used for computing the weights and should not be used as cluster and stratification variables in the analyses and repweights should be used instead (as they include the information about stratification and clustering). However, when using data in long format (for mixed model with timepoints nested within ID), it is unclear to me how to account for this nesting (within ID) using repweights? Would the weights still be applicable?

Thank you

AS
Hi there,

Hi there,

We passed your question to our contacts on the PATH Study Team for feedback. We will be in touch as soon as we can.

Thank you!

Alison

rob
Hi,

Hi,

Was there a response to this question?  I believe it would help with my work as well.  My understanding, albert.ksinan, is that you could run the analysis using the full-sample weights and then iteratively re-run the analysis using each replicate weight (100 of them), one at a time.  Then you could estimate the variance/standard error for any parameter you've estimated using equation 5.2.1 in the user guide with constant, c, specific to the PATH study design for the replication method, in equation 5.2.2.  I would appreciate any further comments on this topic as well.

Thank you.

AS
Hi there,

Hi there,
We are following up with the PATH Study authorities about this question once more. So we hope to hear back shortly.

In the future, you may want to consider emailing PATHDataUserQuestions@Westat.com for related questions.

Thank you.

AS
Westat's answer to question

Please see the response from Westat below.

The VARPSU and VARSTRAT can be used with Taylor linearization for variance estimation if the software being used does not support replicate weights. To use the replicate weights, analysts can copy the same replicate weights to all rows that belong to the same respondent when transporting data from wide format to long format.

albert.ksinan
Thank you

Thank you very much for your response.

Log in or register to post comments