Many were curious to see what the prediction model would have shown if the data were smoothed instead of raw. I have incorporated a 7 day rolling average into the data and prediction.
One way (and to me perhaps the first way) to view the fact that real deaths came in under the center of your range (but clearly within your range) is that the prior data from the Haas report was itself not quite robust enough to get to the exact answer. Clearly, efficacy in preventing mortality is not as high as efficacy in preventing mild COVID, per infection. But is efficacy in reducing mortality was X% (for various X in various age groups), not 96% or 97% or whatever, by the limited Haas data set. Now we see that efficacy in preventing mortality might (1.1)X or (1.2)X in those age groups? But still less than 96-97ish percent.
This really is a level of statistical understanding that even a lot of well meaning data people who got As in undergraduate statistics classes are going to get hung up on. And that's a good reason why absolute risk reduction numbers, or doses/life saved should be expressed in such a paper as the Haas paper as a way of clarifying the results. Of course, those numbers would not have looked particularly good for people under the age of about 50, so...
Now the question: is this totally explained by an epidemic among the unvaccinated? If not then a lot of things we believe about the "end" of the pandemic need revisiting and governments have some 'splaining to do.
My initial take on this would be the model, even as simplistic as it is, is accurate but needs a scaling factor added to lower the curve slightly. That would seem to suggest that the vaccines do lower the death numbers but only marginally. Am I way off here?
One way (and to me perhaps the first way) to view the fact that real deaths came in under the center of your range (but clearly within your range) is that the prior data from the Haas report was itself not quite robust enough to get to the exact answer. Clearly, efficacy in preventing mortality is not as high as efficacy in preventing mild COVID, per infection. But is efficacy in reducing mortality was X% (for various X in various age groups), not 96% or 97% or whatever, by the limited Haas data set. Now we see that efficacy in preventing mortality might (1.1)X or (1.2)X in those age groups? But still less than 96-97ish percent.
This really is a level of statistical understanding that even a lot of well meaning data people who got As in undergraduate statistics classes are going to get hung up on. And that's a good reason why absolute risk reduction numbers, or doses/life saved should be expressed in such a paper as the Haas paper as a way of clarifying the results. Of course, those numbers would not have looked particularly good for people under the age of about 50, so...
Now the question: is this totally explained by an epidemic among the unvaccinated? If not then a lot of things we believe about the "end" of the pandemic need revisiting and governments have some 'splaining to do.
Nicely done, homey. Nicely done.
Thank you. Very helpful.
My initial take on this would be the model, even as simplistic as it is, is accurate but needs a scaling factor added to lower the curve slightly. That would seem to suggest that the vaccines do lower the death numbers but only marginally. Am I way off here?