Hello, not sure whether this qualifies as a bug, but I feel this could be improved.
I recommended to a student to use rineq::ci() in her thesis. The socioeconomic variable was a 14-point index. Now the fractional rank does not handle ties, so the actual CI computed depends on the order of cases in the data file and she and I got different results. I was able to figure out the reason, but she was not. Here is my suggestion:
- Warn against using
rineq::ci() with grouped data OR
- Warn against using
rineq::ci() with grouped data but let ci() randomly order the microdata before computing ranks, with large enough data sets this will help getting close to the "correct" CI OR
- Add option to compute CI from grouped data (prefered)
Best
Hendrik
Hello, not sure whether this qualifies as a bug, but I feel this could be improved.
I recommended to a student to use
rineq::ci()in her thesis. The socioeconomic variable was a 14-point index. Now the fractional rank does not handle ties, so the actual CI computed depends on the order of cases in the data file and she and I got different results. I was able to figure out the reason, but she was not. Here is my suggestion:rineq::ci()with grouped data ORrineq::ci()with grouped data but letci()randomly order the microdata before computing ranks, with large enough data sets this will help getting close to the "correct" CI ORBest
Hendrik