Closed aaronspring closed 1 year ago
I will do that shorty. However, in the future meeting with the other teams we should discuss that "Sem's" way does not seem to be the best... Calculating a percentile threshold out of 20 values does not seem robust. For this reason I will keep in our notebook also the old way of calculating that in a comment section. Then we can fast put that back in, in case we agree with the other teams that this is a better way of calculating this threshold.
agreed to keep to easily switch back. but do you want to stick to calculating the threshold with values from different months? Previously we just did percentile over all samples, i.e. including all months. I think that should be also changed.
Yes that is right! We should implement a rolling window of +-15 days around the day in the year we are talking about.