Code location in evaluate.py
Originally posted by @janetrbarclay in #211 (comment)
I think it's probably worth getting some broader input on this. A higher threshold makes sense for the top / bottom 10% metrics, but maybe not the others?
This change would mostly affect the reach and reach_month metrics (and the new biweekly metric and maybe the year metric). I don't really know the extent of the effects so did a quick data summary below.
There are 907 reach - partition pairs in the observation dataset. 192 have <= 10 temps (and therefore currently would have no metrics calculated) Increasing the cutoff to 20 would leave 256 reaches without metrics.
There are 7685 month - reach - partition pairings. 40% (3624) of them have <= 10 temps. Increasing the cutoff to 20 would leave 52% (4025) without metrics.
Code location in evaluate.py
Originally posted by @janetrbarclay in #211 (comment)
I think it's probably worth getting some broader input on this. A higher threshold makes sense for the top / bottom 10% metrics, but maybe not the others?
This change would mostly affect the reach and reach_month metrics (and the new biweekly metric and maybe the year metric). I don't really know the extent of the effects so did a quick data summary below.
There are 907 reach - partition pairs in the observation dataset. 192 have <= 10 temps (and therefore currently would have no metrics calculated) Increasing the cutoff to 20 would leave 256 reaches without metrics.
There are 7685 month - reach - partition pairings. 40% (3624) of them have <= 10 temps. Increasing the cutoff to 20 would leave 52% (4025) without metrics.