<someone> on Sun, 22 Jan 2017 10:26:24 -0600 writes:
> Dear Dr M?chler, I am using runmed from R's stat
> package. I understand that you are the author of this package.
not of the package - but of function runmed().
I'm reply to R-help, so this answer maybe available to future
web searches.
> I am using the function with even length k=40 and the
> function forces it to be odd as k=41. I am sure there must
> be reason behind this overriding behavior. May I ask if
> you could enlighten me on this? Thank you in advance.
The help page - which one should really read (!) - says that k must be odd.
Why? The median of an *odd* number of observations is "the
middle". That's not the case with an even number, but a very
desirable property, which is kept even when iterating running
medians.
(Further, mathematically there are as many odd numbers as integers, so
odd numbers should be sufficient ;-) ;-))
Martin Maechler
ETH Zurich