Hudson Bay company Lynx-Hare dataset from Leigh 1968, parsed from paper copy http://katalog.ub.uni-heidelberg.de/titel/66489211 (This is the entire Table II)

Leigh1968_harelynx.csv

Dear Mr. Osthege,

I believe there is a typo in this dataset. At year 1862 the hare population should be in a growth period, but it does a sudden dropdown to 5000. We believe that 5000 -> 50000.

I can not open a pull request, but forked you gist and made the changes here. We were using your dataset for a ML project at our university, and we needed to introduce this change for our models to make sense.

Best wishes,
Adrià Lisa.

hi @AdriaCoding, thanks for commenting.
I checked it again and it turns out that I have a typo, but it's not what you pointed at: its Table III not Table II.

The scan available at https://archive.org/details/somemathematical0000symp_o1x7 clearly shows "5,000" hare in 1862.

I do feel sorry for your ML model, but the dataset says 5,000 and if that was a typo, it happened more than half a century ago.

You might want to try a Student-t for the loss or likehood function to get some robustness against outliers.

best
Michael

Thanks for your time Michael.

Indeed I was suspecting that the typo comes from the original authors. Maybe it is not a typo, as the hare population is quite noisy all along the dataset, possibly due to external factors that do not affect the lynx population as much.

Thanks for your suggestions, i will be trying them out. 😄