|
Post by aufan on Mar 11, 2021 20:47:52 GMT
But what is clear is that they are comparing the DAILY growth rate between the periods.
You said that it was the growth over 100 days, which is not anywhere close to what the paper said.
|
|
|
Post by Bevo on Mar 11, 2021 21:21:53 GMT
Change in daily growth rates over a 100 day period.
Isn’t that what it is? And, it’s 2% change in absolute rate.
There was a "growth rate" at the reference period... (which, I guess was actually negative?). And, a different "growth rate" after 100 days... the latter was 2% lower.
If the rate changed from 5% to 3% , they’d be headlining a 40% change in the rate. It’s a 2% change in the correlation coefficients.
|
|
|
Post by aufan on Mar 11, 2021 22:41:43 GMT
It is a 2% change in the daily growth rate. So for your analogy to work about interest rates, 2% daily growth is 2% interest compounded daily, which is equivalent to an 137,640% APY loan. I'm not sure how you get 5% vs 4.9%, but that is not a correct analogy.
For your analogy to be correct, if you had a 5% daily interest rate (or 5,421,184,057% APY) and someone offered you a method to drop that daily interest rate by 2% after 81 days to 3% daily interest (or 4,848,172% APY), I think you would look into that method. I don't think you would dismiss it as nothing.
When talking about exponential growth, small percentages add up big time. A small change in daily growth has huge impacts over the days, weeks and months. I think this is why the direction from experts has been multi-pronged. There is no way to prevent spread (except for complete isolation), so any measures we can do to reduce spread can really add up in the long haul. Even a 2% reduction in your chance to spread the disease to people you interact with adds up over time.
Of course there is no second earth with COVID-19 to run a control, so we will never truly know the impacts. But logically, simple measures to reduce the spread seem to make sense to me.
|
|
|
Post by Bevo on Mar 11, 2021 23:59:55 GMT
The study never showed what the daily growth rate was. As I mentioned, it might even have been negative.
All they said was: the average daily growth rate in reference period was 2% higher than the average daily growth rate in the 20 day period, 80-100 days later. It’s complicated... but, I think you need to rethink what that means. Especially, if you can’t see that 4.9% is 2% lower than 5%.
All this statistical magic is interesting, I guess. But, we now have a lot of empirical evidence to look at. It all leads to one conclusion: government mask mandates don’t stop this virus. They don’t even slow it down in any meaningful way. Maybe voluntary mask wearing slows it down? It makes sense that it should. But, even that depends on what the actual primary mode of transmission is. We still don’t really know. We don’t have any real evidence that mask have made any significant difference.
I prefer voluntary measures. They seem to be equally effective, or ineffective... depending on how you want to look at it. Financially, and socially I think we’d be in a far better place now had we chosen that path. Instead, we STILL have kids not going to school because people are afraid... even when the science says, they shouldn’t be.
I don’t even know how we bring people out of this psychological paralysis. The damage imparted over the past year has been incredible.
|
|
|
Post by bluehen on Mar 12, 2021 0:54:27 GMT
I'm always interested in the dumbth factor. If all pro maskers and all anti maskers were intelligence tested which group would have the highest average IQ ? Don't know, just asking.
|
|
|
Post by Bevo on Mar 12, 2021 1:13:32 GMT
I'm always interested in the dumbth factor. If all pro maskers and all anti maskers were intelligence tested which group would have the highest average IQ ? Don't know, just asking. It’s not a matter or “pro” or “anti” Hen. It’s a matter of figuring out what works, and what doesn’t. And, at what cost? We could all put bags on our heads. That would work. But, at what cost ? It would be great if we could truly show that masks were supremely effective at stopping, or even significantly slowing the spread of Covid. I WISH I could find some data that shows that. I can’t. But, I guess that just makes me stupid?
|
|
|
Post by bluehen on Mar 12, 2021 1:16:17 GMT
Gonna host an old time music jam this afternoon, outside on the deck. We'll be spaced about 4-5 ft apart with masks and most of us have been fully vaccinated. Wow, I wish I was there. Somebody has to put some videos on YouTube.
Not many youtube jam videos of my friends and I but here is one from many years ago, Hero, playing at my place during a picnic/party. We jam like this almost every week when weather permits it outside and safely.
|
|
|
Post by bluehen on Mar 12, 2021 1:24:42 GMT
....and here's another one from years ago , very late night in a friends barn loft :
Mrs Hen , although just out of the video, is providing the bass playing
|
|
|
Post by aufan on Mar 12, 2021 4:27:00 GMT
All they said was: the average daily growth rate in reference period was 2% higher than the average daily growth rate in the 20 day period, 80-100 days later. It’s complicated... but, I think you need to rethink what that means. Especially, if you can’t see that 4.9% is 2% lower than 5%. You are thinking linearly. Yes 4.9% is 2% lower than 5%, but that is the linear relationship. This is an exponential problem.
Lets look at their numbers and do the math, it is actually a very simple algebra problem:
If you use their numbers, you can plug in an arbitrary baseline and see what the result is after 100 days. I'm sure we agree that based on they said, if the baseline is 10%, then daily death growth rates in the days after implementation are as follows:
Days 1-20 is 9.3%
Days 21-40 is 9%
Days 41 to 60 is 8.6%
Days 61 to 80 is 8.4%
Days 81 to 100 is 8.1%
So I also assume we can agree on what daily death growth rate is, and how to use it to discretely calculate cumulative deaths. With a daily death growth rate of 10%, if on day 0 you have 1 cumulative death, that means on day 1 you have .10 deaths for a cumulative 1.1 deaths. On day 2 you have .11 deaths for a cumulative 1.21 deaths. And so on for 100 days. (Technically their method uses continuous growth using natural logs, and this example would be 9.53% growth { 9.53 =[ln(1.1)-ln(1)]*100 }, but its simpler to calculate with discrete growth.)
If the daily death growth rate was steady at 10%, if you had 1 cumulative death on day 0 you would have 13,780 cumulative deaths on day 100.
If the daily death growth rate was based on above values from the study, if you had 1 cumulative death on day 0, you would have 4,117 cumulative deaths on day 100.
On day 100 cumulative deaths are approximately 3.3x higher without the implementation, assuming a baseline of 10%. That is quite significant!
But what about other baselines? A baseline daily death growth of 50% results in ~4.1x10^17 deaths on day 100, but with their calculated reduction in growth it is ~1.7x10^17 deaths on day 100. A factor of about 2.4 cumulative deaths. Though I'm pretty sure since civilization still exists, we weren't at 50%. Obviously the baseline growth rate has an inverse relationship with the impact of 2% (even at negative growth rates), but even at the apocalyptic level of 50% daily death growth rate, a 2% decline still has a significant mathematical impact on the cumulative deaths. We can assume that the death growth rate was under 50%, and the impact over those 100 days would be better than 2.4x fewer cumulative deaths.
The analysis is designed so they can group large data sets together that have varying absolute death rates, so a high population area doesn't wash out low population areas. It is the growth rate change that is significant and what they are after. The bottom line is that 2% decrease in daily death growth rate is quite significant, regardless of what the baseline is.
I know that at first glance, 2% seems like nothing, especially when you see things like "100 days" and intuitively say that is 2% drop over 100 days. But when you realize this is a exponential problem, that it is a 2% drop in daily death growth rates after 100 days, 2% becomes extremely significant.
If you'd like I could probably upload the spreadsheet to google docs or something. But it is very simple to create, just a drag and drop across 101 rows to calculate cumulative values based on a growth rate. Just adjust the growth rate for days 1-20, 21-40, etc. Make the baseline a cell you can adjust to change all the calculations.
I know there are challenges with studies like this because of all of the confounding variables, but strictly based on the math this is a very good showcase that masks are effective for reducing deaths and spread.
|
|
|
Post by Bevo on Mar 12, 2021 17:32:50 GMT
So I also assume we can agree on what daily death growth rate is, and how to use it to discretely calculate cumulative deaths. With a daily death growth rate of 10%, if on day 0 you have 1 cumulative death, that means on day 1 you have .10 deaths for a cumulative 1.1 deaths. On day 2 you have .11 deaths for a cumulative 1.21 deaths. And so on for 100 days. I'm not following you here. If the Daily Rate is growing at 10%, then I see the following: Day 1 1 Death, Cumulative Deaths = 1 Day 2 1.1 Death, Cumulative Deaths = 2.1 Day 3 1.21 Deaths, Cumulative Deaths = 3.3 Day 4 1.33 Deaths, Cumulative Deaths = 4.6 And so on. On Day 100, you'd have 13,780 deaths, and 151,576 cumulative deaths. If we were seeing those kinds of death rates, I'd be wearing 30 masks and NEVER leave my home. In Kentucky, the FASTEST rise we saw in Daily Deaths was in January of this year. At the start, Daily deaths were at 24 per day. By the end of the month, we were at 46. Over that 31 days, our Daily Death rate was increasing at a rate 2%. I even tested this using their definition. ie: Taking the difference between two days of the Natural log of the death rate X 100. It's 2.07% that way. After that, the rate immediately began to fall. As an aside, there was ZERO correlation between either the rise or fall and any changes in mask mandates or changes in indoor dining rules. Looking over Daily Death rates in EVERY state, the fastest increase in deaths I saw anywhere was in Alabama. (Why is Alabama always #1?) That was an increase from 51 deaths/day to 154 deaths/day over a 15 day period. THAT, was an increase in Daily Death rate of 10% per day. Thankfully, it didn't last long. If the improvement from wearing masks were anything close to the levels you calculated, NO ONE would be arguing about whether to wear them, or not. Sadly.. I think the study is showing a 2% decline in regression coefficients, which.. is just a linear percentage. It's linear because they have converted it to a log function. One that is minimal, but technically, statistically significant. At least, in the way the study has calculated it It's a rather complicated way to show the data. That's what people do when they can't find a way to make a convincing case. Make it hard for others to understand.
|
|
|
Post by Bevo on Mar 12, 2021 18:53:44 GMT
Ok... it's taken me awhile, playing with excel, to develop math models that can truly simulate this. I'm starting to become very confident that I understand the numbers.
If, as you say... the actual rates dropped from 10% down to 8.1%, the impact would be TREMENDOUS.
If I assume a baseline of 10% for 100 days, the cumulative # of deaths would be 151,576.
IF I change the growth rates, to what you calculate for each of the 20 day periods, the cumulative deaths would only be 54,490. That's a BIG difference... 64% fewer, and a significant benefit. However, I think we can both agree, the real data just doesn't show that at all... not even close.
If, instead... I assume that the rates drop from 10% to 9.81% (a 1.9% drop), and apply the reported changes over each 20 day period, the cumulative number of deaths drops to 136,715. That's 15,861 lives save, a 9.8% drop. Still, a very a significant number.
However, the actual growth in death rates is NOT 10%, except for very brief periods. Over time, it's a MUCH lower number. That's why they never show it.
If I use the fastest increasing month for Kentucky, 2%..... I get the following:
The Baseline... (2% for 100 days), the cumulative number of deaths would be 319. If I apply the relative % drops in the rate for each of the 20-day periods, as in the 2nd case above, the cumulative deaths drops to 315. That's 4 lives saved, a 1.25% drop. Now, we're starting to see numbers that are closer to what we're actually seeing.
Still, it's a drop in deaths. But, a much smaller number. It's the level of change that is far more likely to "disappear in the noise". Which, is what we see.
As I suspected, the authors of this paper are trying their best to amplify very small changes into something they can use to justify their recommendations. And, of course.. the REAL magic is: How did they adjust the numbers with the other factored regressions...? They don't show ANY of that work.
This, is a junk study. One that was WIDELY reported by virtually every major media outlet as "justification for masking mandates".
But, what do I know? I'm just a low IQ dumbth.
|
|
|
Post by Bevo on Mar 12, 2021 19:02:13 GMT
What I would take out of this is.... Universal masking helps quite a lot when the virus is really out of control and growing quickly (Like, 5-10% daily growth)
Once the growth rate is brought back to something lower, the benefit drops very quickly. Hence, allowing people to decide for themselves was always going to be sufficient.
When people perceive things to be bad, they'll take extra precautions. They'll do this on a minute by minute basis.. Meaning, when they're outside, and away from everyone? Probably take their mask off. When they're in a crowd, they'll likely put it on. When there isn't much virus in their area, they probably won't wear it at all.
That's smarter, and JUST as effective as blanket state-wide mandates requiring everyone to wear masks outside of their homes.
|
|
|
Post by aufan on Mar 12, 2021 19:11:50 GMT
You’re not quite getting it, but you’re almost there.
Think of deaths as the principle, and the daily death growth rate as your daily interest rate. A 10% growth per day means your $1 turns into $1.1 total after a day. It is $1.21 after 2 days. You don’t get your interest AND your principle back each time the interest compounds.
It is the rate that cumulative deaths grows on a daily basis.
I know this to be correct, because they clearly defined the death growth rate as the natural log of the cumulative death rates of a given day, minus the natural log of the cumulative death rates on the day prior.
I also know that because it’s a basic formula to calculate continuous growth. Growth rate = ln(p1)-ln(p0) with p1 your population (or money or deaths) at time 1, and p0 your population at time 0.
You’ve got the correct logic, but you’re compounding the number. Take your calculation and make it not compound and you’ll get there (or an approximation based on discrete growth rather than continuous).
You’ll also see that my calculation is correct, and the impact, assuming their numbers are correct, is quite significant.
|
|
|
Post by aufan on Mar 12, 2021 20:49:32 GMT
In Kentucky, the FASTEST rise we saw in Daily Deaths was in January of this year. At the start, Daily deaths were at 24 per day. By the end of the month, we were at 46. Over that 31 days, our Daily Death rate was increasing at a rate 2%. I even tested this using their definition. ie: Taking the difference between two days of the Natural log of the death rate X 100. It's 2.07% that way.
That was not their definition, nor how you calculate growth rate of a population. You are calculating the growth of the growth of the death rate (hence your really big numbers).
So to do it correctly, you take the natural log of cumulative deaths on a day, and subtract it from the natural log of the cumulative deaths on the previous day, and the result is the percent growth of deaths over that period. Since the period is a day, it is the daily death growth rate.
If there were 100 total deaths yesterday, and 1 person died today, there are 101 total deaths today. To calculate the daily growth in that time period:
ln(101)-ln(100) = 0.995%
Since the time period was 1 day, it is 0.995% daily growth.
|
|
|
Post by Bevo on Mar 13, 2021 16:48:51 GMT
If there were 100 total deaths yesterday, and 1 person died today, there are 101 total deaths today. To calculate the daily growth in that time period: ln(101)-ln(100) = 0.995%
Since the time period was 1 day, it is 0.995% daily growth.
Ok... I understand the math of what you're saying. I'm not sure I understand the logic of it. I guess it smooths out the data. But, the Daily Death Rate is a number... ie; 24 Deaths per day. If the RATE is INCREASING, the number must be bigger the next day (on average) to something like 25 Deaths/per day. So, the rate is, in fact, compounding. And, I can use compounding interest formulas to imitate the data. I like to try to use real world data to test the numbers. I looked at the FASTEST period of growth in Daily Death rates in Kentucky. Jan 1-Jan 31, this year On Jan 1 , The Death Rate was 24.11, the Cumulative Deaths were 2,660 On Jan 31, the Death Rate was 45.78, the Cumulative Deaths were 3,760 I can EXACTLY match that pattern using a compounded interest rate of 2.17%. Using their method, with those numbers, I get a Daily Death growth rate that averages 1.09%. It increases ever day as the cumulative number grows. But back to my point. If they see a 1.9% difference in the Daily Death Growth factor, that HAS TO BE A Relative %. Ie: Some Number * 0.981 It can't be a a difference in the Daily Death Growth rate of 1.9, because the number isn't THAT big to begin with. And, my model of calculation, using compounded interest formulas, can be an accurate predictor of expected deaths, or changes in total deaths. And, that leads back again to: Their study did not see much of a difference, in practical terms.
|
|