The mode, median, and means of a series of number ($x_1,x_2,...,x_N$) can be roughly thought of as the points that minimize the $p$-norm of the sequence for $p\in \{0,1,2\}$. The median is $c=\min_c\sum_i^N |x_i-c|^1$ (ignoring uniqueness of $c$). One can generalize this to get quantiles $\tau$ other than the median ($\tau=0.5$), by making the discrepancy function asymmetric. By scaling up losses on one side of $c$ compared to the other side we get $c_\tau=\min_c (1-\tau)\sum_{x_i\leq c} |x_i-c|^1+\tau \sum_{x_i>c} |x_i-c|^1$. We can make this same generalization to the mode and mean, but do these quantities have nice interpretations or usages?


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.