Panel id NA in ifelse #59
Conversation
so as to avoid "NAs introduced by coercion" warning
actually able to open the large panel files
|
The basic identification had been using V1023, which reports to area type (capital, rest of the metropolitan region, etc.). that is useless to identify individuals within a household. After checking out the Stata code, I realised the correct variable to use here would be V2003, which order number within the household. V2003 is also used in the advanced identification. This mistake was probably why we were getting worse results with the advanced identification than with the basic one - the latter was less strict than we expected. I am unsure, however, if including V2003 in .by arguments will be enough to solve the problem, as Stata does something more complex with it |
|
The current solution paste0(V1014, id) does not work because individual 23 from panel 1 will have the same id as individual 3 from panel 12. Using as.hexmode was suggested. It should work until panel 16 is created. Then, we will need to come up with a new solution. I tried implementing it and ran tests on panel 2. % found went from 60%ish in interview 5 to 55%ish. Worse than that, the % found now grows as interviews progress, which makes no sense. Therefore, I am reversing the last commit |
This reverts commit 5d3613e.
Adds NA check so as to no throw warnings when paste0 on panel number and id.
Tested and working.