When you train on old data, you get a lag of about 10-12 days between changes in cases and corresponding changes in deaths. There are several reasons that could be not true on new data:
1. Cases are getting caught earlier by more/faster testing.
2. Cases are leading to fewer or slower deaths (due to either treatment or population effects)
3. The lag on old data is using the reported date of death, but that’s not the same as the date of the reporting of the death, which has an additional lag.
I suspect a couple of things might be worth considering, but I’m not the expert here either so take everything with the view I am speculating/thinking aloud not stating any findings.
I don’t think testing will tend to lower the CFR as that testing will move things towards the real IFR rather than the CFR. This probably related to point 1 & 2 above.
I think the 10-12 days from the old data to say we see movement in the death data due to the new cases probably has some type of skew in it, the older the data the more likely it will be complete. That should be driven by the the death reporting distribution (and perhaps even corrections). The closer the old data gets to the new threshold of new deaths it should under report due to the lag. Perhaps we need to look at the distribution of reported deaths over that 8+ week period before trying to assess the results after the 10-12 days. I’m not sure if that is what you are saying in point 3.
When you train on old data, you get a lag of about 10-12 days between changes in cases and corresponding changes in deaths. There are several reasons that could be not true on new data:
1. Cases are getting caught earlier by more/faster testing.
2. Cases are leading to fewer or slower deaths (due to either treatment or population effects)
3. The lag on old data is using the reported date of death, but that’s not the same as the date of the reporting of the death, which has an additional lag.
Are you saying it’s (at least partly) #3?
I suspect a couple of things might be worth considering, but I’m not the expert here either so take everything with the view I am speculating/thinking aloud not stating any findings.
I don’t think testing will tend to lower the CFR as that testing will move things towards the real IFR rather than the CFR. This probably related to point 1 & 2 above.
I think the 10-12 days from the old data to say we see movement in the death data due to the new cases probably has some type of skew in it, the older the data the more likely it will be complete. That should be driven by the the death reporting distribution (and perhaps even corrections). The closer the old data gets to the new threshold of new deaths it should under report due to the lag. Perhaps we need to look at the distribution of reported deaths over that 8+ week period before trying to assess the results after the 10-12 days. I’m not sure if that is what you are saying in point 3.