Statement for the record:
So there is no mistaking my intent, everything that follows is based on San Diego County, and may
or may not be applicable to other locations in the state or country.
Bottom Line Up Front:
So here we are, three weeks into our Goober's lastest lockdown of San Diego, and the result is...
He deserves to be recalled. Seriously, that's the kindest thing I can say. Words like "tar and feathering" do come to mind, but I will try to stay reasonable and civilized (not that his side of the political spectrum worries about that). His lockdown has accomplished what he really wanted. Namely the economy is once again jerking spasmodically to a halt as more and more business close their doors forever and unemployment claims are again on the rise. His actions are understandable from that point of view, but not from any other, and particularly not from the viewpoint of a sensible review of the data. Now let me defend those statements.
Short review:
I originally claimed (here) that the Goober was using metrics defined in terms of the original beliefs about the data, which are simply not relevant to today due to improvements in our understanding of the data. I then joined our Goober in doubling down (here). He insisted that we needed to further choke our businesses, while I insisted that that was somewhere between insane and criminal (oh, and I used the word 'malfeasance' in one context and I will further support that claim).
Update:
After one week I reported the score as "Goober: 0, Me: 1". I've now got two more weeks in and I can say the score is now Goober: 0, Me: 3. I can also add this new score as well San Diego County Health Officer: 0, Me: 1 (maybe 2). So far it really hasn't been a close contest. But let's look beyond the standings to see the box score, to show just how badly they've done.
So far my predictions (which are based on my admittedly ad hoc method to predict the deaths) were that the deaths would not see a major rise, and would remain well below the amount the Goober had shown himself to be willing to tolerate - due entirely to the case rise being of a completely different demographic distribution than the original case distribution. That number is around 7 / week / 100,000 population, which is about 234 / week for San Diego County. I will freely admit that I think that is bigger than I would want to tolerate, I'd be seriously concerned if we topped 150 / week, and would think some initial actions might be appropriate once you reach 120 / week (or about 1/2 his number). However, my predictions for the total deaths for the weeks ending July 11, 18, and 25 were 66, 61, and 62, respectively (just under 2 / 100,000). These are comfortably below the point where I'd start to worry (about 1/2 my point of concern). In fact, the actual death totals were 35, 56, and 55. Even smaller than I'd predicted, and way less than the kind of numbers that should cause our Goober to push for re-closing the re-opened business.
I also complained about our County Health Officer (and everybody else) intentionally failing to ask during their contact tracing if the new cases had attended any of the 'protests', which were so vigorously defended by our Goober as protected free speach under the First Amendment. I don't see that the Health Officers and contact tracers had the authority to not ask about attendance at such events. I did a little web surfing and found this quote (which comes from the California Code, Health and Safety Code - HSC § 120175)
Each health officer knowing or having reason to believe that any case of the diseases made reportable by regulation of the department, or any other contagious, infectious or communicable disease exists, or has recently existed, within the territory under his or her jurisdiction, shall take measures as may be necessary to prevent the spread of the disease or occurrence of additional cases.That seems very clear to me. Now what is Malfeasance? Here the first web search led me to this definition
Intentional conduct that is wrongful or unlawful, especially by officials or public employees. Malfeasance is at a higher level of wrongdoing than nonfeasance (failure to act where there was a duty to act) or misfeasance (conduct that is lawful but inappropriate).Having intentionally not asked about attendance at the 'protests' is, as I read it, a clear intentional failure to do their official duty. I further claim that as that intentional failure then led to them being unable to properly determine the source of the spread, and thereby allowed our Goober to order the re-closings, which then resulted in harm to San Diego County residents and businesses. Hence, I claim it is a clear act of malfeasance. Further, our County Health Officer has recently begun to include in their briefings, on the plot of daily new cases, an annotation showing the dates for the incubation period for bars and restaurants re-openings. While I concede that they have shown this in technically accurate manner, it is also in what must be considered an intentionally misleading one in at least two ways. First, it shows the incubation period for the initial reopening, but does not extend the incubation period for the entire time from when the restaurants and bars were reopened until the Goober re-closed them. Secondly, it does not show the relevant incubation periods for the various protests.
I have reproduced my plot from the last post in Figure 1, but have made the change that the data are now 1 week averages, with the average assigned to the last day of the one week period. I also colored two points red, just to make orientation easy. These two points are June 12 and July 4. The other schematic lines are explained below.
Figure 1. Plot of the 7 day running average of new cases for San Diego County along with schematic representations of the two cases. Case 1 in green, Case 2 in yellow, see text for full descriptions of the two cases. The two red points are for the dates July 4 and July 12 and are colored differently only to allow easier identification of various time periods.
I can predict what should happen to such a plot based on the modeling experience I gained earlier. Increased spread will not be seen in the case data for something like 7-14 days. This is due to the increase having an induction period where the cases spread and then re-spread, this is then further delayed by the time lag inherent in diagnosing the cases, and finally another apparent delay due to the averaging reducing the response of the curve. This very slow grow will be followed by a period of essentially exponential growth, which will then eventually roll over (typically showing the same kind of lag as the increase did). We shall now consider in the two most likely cases.
Case 1: If the case rise is primarily due to the reopening of the restaurants and bars, the initial rise would be first visible some where around 7-14 days after the initial re-openings, which occurred May 21, so somewhere in the May 28-June 4 time period, maybe slightly later. This would be followed by an essentially exponential period which would have no reason to roll over until the interaction was reduced (re-closing happened July 7). The reduction would first be visible after a lag of over a week (for the same reasons we see the increases lag). This means the rollover would only have started about the 14th (maybe slightly later), and be pretty flat starting about now. This profile does not match the observed data.
Case 2: If the case rise is primarily due to the 'protests' (which occurred primarily from about May 29 - June 13, with a few smaller ones about a week later), then we'd expect to see an initial rise starting in the June 5-12 time frame, really picking up for a week or two, but then rapidly flattening due to the interaction being reduced to the previous level, but with a higher case number (and a concomitantly larger new case rate), which would appear as a flat zone starting around July 4. This is almost exactly what is observed in the data.
Anyways, like I said earlier, San Diego County Health Officer: 0, Me: 1.
Further Considerations For Predicting the Expected Deaths:
I could write many many pages about what I've done with the data I now have (three weeks of day by day deaths, broken down into age groups of 10 years, and almost five months of daily new case data broken down the same way). I will not inflict that on you dear reader. Instead I will simply state the best way I found to predict the deaths is to use each age group separately and for each of these to use the new cases over one week, to predict the deaths in the next week. The current best fits are by using the death rates as given in Table 1. I will note that using older case numbers like 2 or 3 weeks previous work OK, but not quite as well. If, however, you try to use the aggregate cases, without regard to the age grouping, the predictions are essentially worthless.
__________________
Table 1. Best fits for death rates for each age group using cases from the weeks ending 6/27, 7/04, 7/11, and 7/18/2020 to predict the deaths for the weeks ending 7/04, 7/11, 7/18, and 7/25, respectively. *- There have been exactly 0 deaths for these time periods for these groups, even though there have been 3 and 4 deaths total for these two age groups over the entire CoViD-19 outbreak.
| Age group | Death rate |
|---|---|
0-10
|
0
|
10-20
|
0
|
20-30
|
0*
|
30-40
|
0*
|
40-50
|
0.0055
|
50-60
|
0.0138
|
60-70
|
0.0444
|
70-80
|
0.117
|
80+
|
0.269
|
Table 2. Predictions, both the previous ad hoc ones and the new ones based on better analysis,and the observed deaths.
| Week Ending | Old Prediction | New Prediction | Observed |
|---|---|---|---|
7/04/2020
|
33
|
54
|
27
|
7/11/2020
|
66
|
49
|
35
|
7/18/2020
|
61
|
50
|
56
|
7/25/2020
|
62
|
51
|
55
|
8/01/2020
|
-
|
53
|
??
|
No comments:
Post a Comment