Status:Closed    Asked:Jul 21, 2014 - 11:24 AM

Some of my PERWTs are 0.0? Are my analyses invalid?

I have 650,000 cases in my data set. Only 200 have PERWT of 0.0 but unfortunately a portion of these 200 are my subjects of interest (transracial adoptees, where the race of the individual does not match the race of the household head). I assume that transracial adoption is such a unique situation that these cases do not represent even 1% of people in the US however my concern is that they do exist yet are not part of my weighted analysis due to their 0.0 weight.

In short, is my analysis of transracial adoptees going to be invalid since a few of them do not have a weight?

Do you have the same question? Follow this Question

Staff Answer




You are correct that the individuals with zero weights will not be included in your weighted analysis, however your analysis of respondents with zero weights is not invalid. The weights and samples are generated by the Census Bureau to represent the 100-percent totals for certain data groups, and a weight of zero just indicates that the individual was "over-represented" in the sample. In the contemporary samples, zero weights should only appear in 1990 and 2000 samples. IPUMS-USA also created 1% unweighted samples for both 1990 and 2000. Because these samples are only 1% samples, you will be less likely to find members of very specific groups such as transracial adoptees, but you could compare the figures you find to those produced by the 5% file.


Jul 22, 2014 - 10:05 AM

Report it


IPUMS Abacus beta: Is it possible to recode age to custom groups? such as 1...
Expected release of 2013-2017 5-year data?
Follow-up to question on CWS 2017 release
Encountering errors when running data sets on SPSS
Using IPUMS-DHS data. How to identify strata with single sampling unit?
Login   |   Register

Recently Active Members

View More »

Share |