Estimation Review for 5/13 Estimates
The estimates made on 5/13 are now complete so it’s time to review how they did. Error is calculated using “Balanced relative Error (BRE)”.
This set of estimates was more accurate than the first set. Average error on the daily estimates was 24%, median was 17%, and 67% of the estimates were within 25% of actual (Pred(25)). On a cumulative basis, the average daily error was 4% and median was 3%. 100% of the estimates were within 10% of actual and 67% were within 5% of actual.
I think two things can be improved:
1) I want to tighten up the day-to-day variation. There’s some day-to-day variation across the week in terms of how deaths are underreported or overreported on certain days, and I want to account for that.
2) I want to change to a 13-day lag instead of a 12-day lag. As you can see from the second chart below, if I had used a 13-day lag for this set of estimates they would have tracked even better than they did.
New Estimates as of 5/25/20
This set of estimates uses the 13-day lag and the day-of-week adjustments.