24 Aug
2020
24 Aug
'20
6:06 a.m.
Hi, Boost offers convenient methods in the accumulator library to get e.g. a value (let's say 1234) corresponding to a certain quantile (let's say 0.995). For my specific use-case I wonder if the opposite (value input=1234, output=0.995) is also possible. Is there an elegant way to calculate/estimate the percentile of value in an online fashion (=while accumulating) and without storing all values? One thing I came up with is using extended_p_square_quantile with many probabilities and then do a binary search over the values output by the quantile() function. Is there an easier way to directly get the percentile from the value that I am missing? Best, Timo