search

adamaltmejd / covid

MIT License
88 stars 14 forks source link

False plateau in forecast #24

Closed davgussnavgus closed 4 years ago

davgussnavgus commented 4 years ago

In the current scenario with decreasing numbers of death your forecast is constantly overestimating the numbers pretty severely (~a factor of 2). Its a bit funny since it gives a false impression of a plateau just as FHMs numbers did when the numbers were increasing,

adamaltmejd commented 4 years ago

The model is conserevative and simple by construction, since it is just the historical average lag. If you want to read a more serious attempt at accurately nowcasting the number of deaths, see our working paper here: https://arxiv.org/abs/2006.06840. I might implement the model from thee paper on the website at some point when I have time, but it wont be for a couple of weeks.