I got an answer from a BJS statistician and thought I would share it in case anyone else is interested in the topic of correct settings for analyzing NCVS incident level data. The key lines below are the svyset commands.:

The svyset code in Appendix F still applies, regardless of if you are running tabulations or regressions. If you have not seen it, the NCVS User Guide from which Appendix F is taken might be helpful to you and can be found at:

https://bjs.ojp.gov/sites/g/files/xyckuh236/files/media/document/ncvs_variance_user_guide_11.06.14.pdf. Section 2 goes into more detail about how to get your data prepared, though it too uses language regarding counts/rates, etc. As for an example, here is how you would set up your svyset to run regressions:

* Create yr_grp for svyset *

recode YEAR (1992/1996 = 1)(1997/2005 = 2)(2006/2015 = 3)(2016/2022 = 4), generate(yr_grp)

* Create svyset *

egen strat=group(yr_grp V2117)

svyset V2118 [pweight=SERIES_IWEIGHT], strata(strat) vce(linearized)

V2117 is the pseudostratum variable and V2118 is the sample code variable which are needed to account for the NCVS complex sample design. Once you have svyset your data, you call the svy command and specify that you want to run a logistic regression with “svy: logistic” then you would enter whatever variables you want into your regression. This will account for the complex survey design of the NCVS. As you can see, this remains reflective of the commands in Appendix F. The PSU and strata variables do not change. What would change would be the weight you elect to apply.

The weight that you want to use would depend on your outcome variable. Victimizations and incidents can be examined using NCVS data and more information about the difference between these estimates is included on page 2 of Criminal Victimization, 2022. Details about all NCVS final weights and replicate weights are available in the NACJD documentation archived with the public-use data files. For the 2022 NCVS data, this information is included in the User’s Guide.