Simple stats question that I can't find an easy way to google. Suppose I have a survey question, and have the median response, the number of people who answered, and the total number of people in the population (say 1.1, 15, 200), but not the sd. Can I calculate an error bound?
-
Show this thread
-
Replying to @marquezxavier1 reply 0 retweets 2 likes
-
Replying to @turrible_tao @marquezxavier and
Error bounds on a median depend on the variance in the data set. So if something about the variable itself could let you bound the variance (say, it only goes from 1 to 5) then you could bound the median, but probably not by much
1 reply 0 retweets 3 likes -
Replying to @nickchk @turrible_tao and
booooooooooooooootstrap probably
2 replies 0 retweets 3 likes -
Replying to @eigenrobot @nickchk and
to bootstrap you need the individual data points, but it sounds like we only have some summary statistics instead
3 replies 0 retweets 3 likes -
Replying to @alexpghayes @eigenrobot and
Sorry to butt in but a related question: suppose instead of just P50, you had P50 and P90, and wanted to use those to get the full CDF. Any way of doing so? So far I’ve seen people use those two pts and fit a shifted gamma or poisson mixture but not sure how robust that’d be.
2 replies 0 retweets 1 like
you could generate a bunch of datasets and do some robustness checks along those lines maybe?
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.