Question regarding data discrepancy in the same extracts

USA
1

Hi,

I was facing a question regarding what can lead to huge data discrepancy when you resubmit a previous extract. I extracted data number 24 on October 1, 2021. I needed the data again so I recently resubmitted it again . However, the statistics that I get from the data are different from what I obtained when I initially extracted the data 24. So I went ahead and extracted 24 again without the resubmit option, but the revision option (but not actually revising anything). The data extract number I get is extract 30 in my account. This time around my estimates were exactly same as what I got before. In sum , the statistics that I obtain from extract 24 is different from what I obtain from extract 30 (which is just a revision of 24, without actually revising anything) . For instance, if I do the summary of the variable inctot, the lower bound for extract 24 is 0 while for extract 30, it goes in the negative. I wanted to ask what can lead to this discrepancy, or is there something specific to my ipums account. Thank you.

2

This is definitely unexpected behavior. We are looking into the issue and will provide an update as soon as we have more information. My apologies for the inconvenience.

3

Iā€™m posting this on behalf of Kari since she is out of the office this week.

Thanks for your patience while we looked into this issue. This issue resulted from a rare interaction between how our extract engine is configured and the StatTransfer process for creating a Stata-formatted data file. The issue should not affect other data users going forward; however, you will need to use your revised data extract instead of the resubmitted data extract to get the correct values. To confirm, your extract 30 (the revision), which includes a broader range of values for INCTOT is the correct version of the data.

1 Like
4

Thank you for the confirmation!

1 Like