I think that I would go with a list if there were more than one mode value. If you think about it taking the median would work against the most common occurring because it is likely that the median of the two most common values may not occur as a value at all! At least the list result retains the information of values most likely to occur.
Cheers, bob > On Jul 24, 2020, at 11:46, Raul Miller <[email protected]> wrote: > > https://en.wikipedia.org/wiki/Mode_(statistics)#Uniqueness_and_definedness > > "Finally, as said before, the mode is not necessarily unique. Certain > pathological distributions (for example, the Cantor distribution) have > no defined mode at all." > > That said, just as we can redefine median to be the mean of the two > median values when the length of the sequence is even, we could > redefine mode as the median of the candidate mode values when there is > more than one "most frequently occuring value". > > Thanks, > > -- > Raul > > On Fri, Jul 24, 2020 at 2:36 PM Devon McCormick <[email protected]> wrote: >> >> Hi - I've started reading "Fun Q" which is a book on machine learning using >> the q language. Early on, the author points out that his "mode" function - >> where "mode" is stats-talk for "the most frequent observation" - is >> order-dependent. >> >> I checked my own "mode" and found that this is true of mine as well: >> mode >> ~. {~ [: (i. >./) #/.~ >> mode 1 2 2 3 3 >> 2 >> mode 1 3 3 2 2 >> 3 >> >> This might be an ill-defined statistical concept but does anyone have any >> insight based on practice? Is this order-dependence just a weakness of the >> definition of "mode"? >> >> I could not find "mode" defined in any of the J standard libraries. >> >> Thanks, >> >> Devon >> >> -- >> >> Devon McCormick, CFA >> >> Quantitative Consultant >> ---------------------------------------------------------------------- >> For information about J forums see http://www.jsoftware.com/forums.htm > ---------------------------------------------------------------------- > For information about J forums see http://www.jsoftware.com/forums.htm ---------------------------------------------------------------------- For information about J forums see http://www.jsoftware.com/forums.htm
