Venkatesh Rao@vgr·Nov 10, 2014Replying to @pmarca@pmarca nice thing about bullshitting with median instead of mean is that you have excuse for not high-end explaining outliers.239
Sonal Chokshi@smc90·Nov 10, 2014Replying to @vgr@vgr @pmarca jokes aside, wouldn't mode be the best measure of central tendency here?1
Venkatesh Rao@vgrReplying to @smc90@smc90 @pmarca Nope. Mode can be non-unique, causing ambiguity here. Choice of median vs mean is about expected tail-skew effects.1:07 AM · Nov 10, 2014
Rahul Pathak@rahulpathak·Nov 10, 2014Replying to @vgr@vgr you should publish both right? That makes the distribution skew more transparent?1
mobiusbobs@mobiusbobs·Nov 10, 2014Replying to @vgr@vgr @smc90 @pmarca a distribution may need more than one parameter to describe its characteristics, skewness, kurtosis, up to the Nyquist.