Has anyone ever applied cook's distance method on fisheriris/wine/glass or similar datasets for outlier detection? Did it work? How?

load fisheriris; X = double(meas(:, 2:4)); y = meas(:,1); mdl = fitlm(X,y); k=mdl.Diagnostics.CooksDistance; find((mdl.Diagnostics.CooksDistance)>3*mean(mdl.Diagnostics.CooksDistance));

[William Mueller]

There’s an application in econometrics here:

https://www.mathworks.com/help/econ/examples/time-series-regression-iii-influential-observations.html

Answer 1

[Bernhard Suhm]

You can find an example for using Cook's (and related) distances to pick out unusual years in the financial history here. Not sure how much that'll help you with your project.