I am interested in appending 5 years of data into one single file in order to look at differences in hours worked across years various occupations. I am worried about running into issues with with the SEs when conducting person-level weighted regression analyses in this single file. Is there a suggested way to handle appending multiple years into one when planning to do individually weighted regression analyses and ensure this does not influence the SEs?
IPUMS allows you to add multiple samples to one extract/data file so that you don’t have to append multiple files together (unless you are doing so because you have limited computing power). See this short video tutorial on using the extract system to see how to include all the samples you want to use in your IPUMS USA extract.
When working with IPUMS data from multiple years, you may be interested in time series analysis (studying change over time) or pooling data from multiple years together (to increase sample size or estimate a pooled average or frequency over several years). IPUMS data allows both types of analysis. IPUMS provides replicate weights for ACS samples, which the Census Bureau recommends using to estimate standard errors accurately. For non-ACS data from IPUMS USA, you should use PERWT (person-level weight) or HHWT (household-level weight) for most analyses.
If you are estimating a frequency or other count-related estimate with pooled data (the estimate includes data from more than one sample), you need to modify your sampling weights by dividing by the number of pooled samples.