Which weights should I use to analyze the ATUS Eating and Health Module and the Well-being Module simultaneously?

I would like to analyze data from the ATUS Eating and Health Module, along with data from the Well-being Module (in the same statistical models). Should I use wt06 (along with the corresponding replicate weights of rwt06_*)?

Generally speaking each weight variable is designed for a specific set of variables. In the case of the Eating and Health module and the Well-being module, a bit more information is necessary before answering this question. This is because none of the Eating and Health module questions overlap with the Well-being module questions. In particular, the EHWT variable is available in ATUS samples from 2006-2008 and 2014-2016. The WBWT variable is available in ATUS samples from 2010 and 2012-2013. This raises the question, how are you planning on including data from both of these ATUS modules in one regression?

Apologies for never responding to JeffBloem’s comment! I have returned to this issue recently. To answer JeffBloem’s question, I am interested in including a variable related to general health in a regression model. This variable is included in both the Well-being and Eating and Health Modules for several years. One possibility that occurs to me (for how to analyze these data together) would be to create a new weight variable with values of EHWT for Eating and Health Module cases and values of WBWT for Well-being Module cases–and do something similar for each of the replicate weights. Would this be an appropriate method? Many thanks for any assistance IPUMS Staff can provide.

I’m following up here (again) to post a relevant link that I just came across, in case this is helpful for others: https://www.atusdata.org/atus/resources/linked_docs/Training_Materials/Exercise_Combining_Module_Data_Exercises.pdf

This exercise suggests using the technique I mention above, at least for EHWT and WBWT. Perhaps this could work for the replicate weights, as well?

I’m glad to hear that you found the exercises helpful! The technique you mention above is the correct way to incorporate multiple module weights. When using data from ATUS modules, researchers should pair that module’s weight with respondents to that module. When including respondents from multiple modules, researchers can use multiple module weights, pairing respondents to their respective module weights. The easiest way to do this is to create a new weight variable that takes on the value of the appropriate weight for each respondent. The same rules should be applied for module replicate weights, including RWBWT for person-level analysis and RAWBWT for activity-level analysis of the well-being module.