ProjectMOSAIC / mosaic

Project MOSAIC R package
http://mosaic-web.org/
93 stars 26 forks source link

query about MAD #691

Closed nicholasjhorton closed 5 years ago

nicholasjhorton commented 6 years ago

I've always thought that MAD was "median absolute deviation from the median". But the MAD() function in mosaic is the mean of the deviation of each observation from each other observation. Is there a citation for the current mosaic defined statistic?

rpruim commented 6 years ago

It is used in Tintle et al chapter 9. That's the reason for creating it. I don't know whether they have a reference to other places it is used. The help file does mention stats::mad(), which does the thing you were expecting, but we can perhaps mention this issue more explicitly.

rpruim commented 6 years ago

It would be possible to move this function into the ISIwithR package if thought it was not generally useful to our primary audience.

rpruim commented 6 years ago

I've put this in the documentation:

All pairs mean and sum of absolute differences

Description

The functions compute the sum or mean of all pairwise absolute differences. This different from stats::mad(), which computes the median absolute difference of each value from the median of all the values. See the ISIwithR package (and the textbook it accompanies) for examples using these functions in the context of simulation-based inference.

nicholasjhorton commented 6 years ago

This looks great. Only suggestion: “different” should be “differs”.

All the best,

Nick

On Feb 17, 2018, at 9:47 AM, Randall Pruim notifications@github.com wrote:

I've put this in the documentation:

All pairs mean and sum of absolute differences

Description

The functions compute the sum or mean of all pairwise absolute differences. This different from stats::mad(), which computes the median absolute difference of each value from the median of all the values. See the ISIwithR package (and the textbook it accompanies) for examples using these functions in the context of simulation-based inference.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or mute the thread.

rpruim commented 5 years ago

This was fixed long ago. Closing.