Hello everyone,
I have some questions regarding using the sensatt command to search in a systematic way for and identify the characteristics of would would be a 'killer confounder'.
The creators of the sensatt command have written an article where they use sensatt to test the sensitivity of the results on real world data (https://ftp.iza.org/dp2149.pdf). In this article they have created two tables (table 9 and 11, page 54 and 56) where they incrementally increase "d" and "s" by 0.1 from 0.0 to 0.7, which makes it easy to see when the results starts to breakdown due to the confounder.
What I can not understand is how to create "d" and "s" in such a way that I know the increment will be 0.1 of one while keeping the other fixed. The authors writes that one can pretty much set the probabilities for p11 and p10 to what ever value you want as long as they are they same since "d" = p01-p00 and "s" = p1. - p0. But two problems arise from this:1. I can set p01 and p00, but p01 - p00 is NOT the exact difference between the two. 2. its not possible to set p1. or p0. , so from which values are p1.-p0. calculated?
I would very much appreciate if anyone had any input to give on this issue, thanks in advance.
regards
/Anders
I have some questions regarding using the sensatt command to search in a systematic way for and identify the characteristics of would would be a 'killer confounder'.
The creators of the sensatt command have written an article where they use sensatt to test the sensitivity of the results on real world data (https://ftp.iza.org/dp2149.pdf). In this article they have created two tables (table 9 and 11, page 54 and 56) where they incrementally increase "d" and "s" by 0.1 from 0.0 to 0.7, which makes it easy to see when the results starts to breakdown due to the confounder.
What I can not understand is how to create "d" and "s" in such a way that I know the increment will be 0.1 of one while keeping the other fixed. The authors writes that one can pretty much set the probabilities for p11 and p10 to what ever value you want as long as they are they same since "d" = p01-p00 and "s" = p1. - p0. But two problems arise from this:1. I can set p01 and p00, but p01 - p00 is NOT the exact difference between the two. 2. its not possible to set p1. or p0. , so from which values are p1.-p0. calculated?
I would very much appreciate if anyone had any input to give on this issue, thanks in advance.
regards
/Anders
Comment