search

tnc-br / ddf-isoscapes

4 stars 0 forks source link

Run Oxygen Pipeline T-test #18

Closed jmogarrio closed 1 year ago

jmogarrio commented 1 year ago

Evaluate the current pipeline using non-averaged data

benwulfe commented 1 year ago

@gretaabib I think there are several concrete suggestions for this T-test:

a) It should live in the API (probably in the future, earth engine) Repository where we would otherwise use for production. This allows us to run tests closer to production.

b) It should leverage Kader's code, which is clean and also aligns with production use case.

c) It should run on all elements individually

d) It should run on the combined elements

e) It should run on all proposed isoscapes in a consistent way to enable comparions.

f) It should use 'glossary' naming conventions

Dr. Martinelli also would like the result expressed as a "probability of fraud".

jmogarrio commented 1 year ago

Clarified with @gretaabib that this is done, but filed #96 to track continued improvement on the performance of the t-test.

Maybe the last thing we want here is a for the results we originally obtained, in addition to steps for reproduction. Both of these should be straightforward to obtain now that running the Colabs is more streamlined.

Still, at this point I think that warrants a separate tracking item. Since @gretaabib is working on the improvements in #96 I'll follow up with documenting the initial results in #97, and close this issue out.

I'll also transfer @benwulfe's comment to #97.