COVID-19 Daily Update for Yesterday, Tuesday, 8/25/2020

Is it Signal, or Is it Noise? (World Wide Trends Update)

We discussed these two charts a few days ago. But here's the update: 

The question at the time was whether the seeming uptick in the 7 day average of the daily change in new cases seen in the 7 day average was signal or noise, given that the 14 day average was still going down. 

As of today, the 14 day average is also going up. So for now at least, it appears that this metric is getting worse. 

However, it appears that much of this is driven by the change in recovery rates. The number of daily new cases appears to be slowly declining:

First Confirmed Case of Reinfection (Is that as Scary as it Sounds?): 

The first truly confirmed case of reinfection has now been reported from Hong Kong. Previous reports of reinfection may have been due to false positives, or to a test picking up on dead viral material, or from a resurgence of the previous infection. 

However, in this case, we know it's real. The RNA of the virus was sequenced from both infections, and it is clear that the second infection was both real, and different from the first. 

I have seen quite a few news stories, and/or posts on social media with people freaking out. 

What does this really mean? And should we be panicking? 

First, it doesn't mean that the virus has mutated enough to reinfect people. The mutations were enough to use RNA tests to show this was a different strain, but we have been tracking mutations of the virus, and using those to track the family tree of the pandemic from the start. There is no evidence (yet) that the virus is mutating in ways that help it to avoid immunity (like the flu does).

Instead, it would appear that the man's natural immunity simply decreased to the point where he could be reinfected. This is how most other corona viruses (like the corona virus versions of the common cold) end up re-infecting us. Meaning this was expected. The reinfection took place after four and a half months from the time of initial infection, meaning immunity likely lasts at least that long, and most likely longer on average, since this was the FIRST confirmed case like this so far. 

That we lasted this long is GOOD news. 

What is even more exciting was that his second case was completely asymptomatic. Meaning, that his immunity wasn't actually gone. While it didn't prevent him from being reinfected, it DID prevent him from getting ill with ANY symptoms the second time through. 

This is also good news, this is VERY good news. 

We also know that many of the vaccines currently in development appear to be producing a stronger, and longer lasting immune response than the disease itself. 

While this does mean that natural herd immunity is a losing strategy, we already knew that. By the time we could conceivably anything close to herd immunity levels of population immunity... that immunity would be going away for many, and things would just start over. Although the disease would likely be less deadly to may the second time around, this would make it impossible to prevent the elderly and/or vulnerable from catching the disease at least once, and that would mean an unacceptably large death tole. 

However, with a vaccine, things look very different. 

While this case of reinfection may mean that we will need a booster at 6 months, it does not mean that a vaccine would not work. And, with a vaccine, we could actually reach (and maintain) herd immunity levels such that we could allow people who are elderly and/or vulnerable to go back out into society with minimal risk of getting infected. 

The news articles sound scary... but I believe that this isn't actually bad news. 

COVID-19 Daily Update: Automation Progress Update:

Today is a model discussion day. I will give a pandemic status update tomorrow. 

I have been working recently on automating the process of gathering data every morning. Frankly, I'm surprised it took me this long. 

Previously I had been moving data to my spreadsheet from various sources by hand. For those interested, here's a link to my spreadsheet, with all my old data. 

As nice as it was to have all the data (and figures) in one place, where everyone could look at it, and intuitively see how I did any calculation... as I began to track (and correlate) a larger set of data, this became unwieldy, and simply took too much of my time. 

I spent yesterday and this morning writing python code to automate this process. 

For those interested, the code is available on github: https://github.com/jlc42/JLC-COVID-19-Tools

This is a very preliminary work in progress. Currently, it has scripts to gather data which I ca run each day. I still need to write some code to automatically parse some of the web pages I scrape. But at least the data is saved. (Some date, notably from Georgia and Texas goes away if you don't scrape it daily. Some of that data is saved each day by CovidTracking, but other bits are simply lost if I don't scrape it myself daily, namely the testing data on antigen tests or serology tests. Texas puts that in their spreadsheet, but only has the "daily" value for the antigen and serology tests there... so yesterday's numbers are 'wiped out' by tomorrow's numbers.)

I can now put that on a chron job, and have it run every night while I am sleeping, and the data will all be saved. 

Next I need to write some code to go through all the saved data, and organize it into the useful fields I track... that's a larger task, that will take a LOT of time. For NOW, I have it pulling down (and organizing) the US state data from CovidDTracking... and then building a single plot... 

So... after two days of work... drum roll please.... here's my single plot, automatically generated for me this morning: 

And there it is, in all its glory. Here's the same figure plotted from my spreadsheet: 

I happen to think the scripted python version looks better. 

Side note: things ARE improving in New Mexico, yes Tests are down along with cases, but that is because the DEMAND for tests is falling, as true infections fall... you can tell because the % of the tests that are positive is also falling. 

Back to automation... 

This single figure may not look like much, HOWEVER, with a simple loop, tomorrow I will be ready to produce this figure for EVERY state in the US. 

And with a little bit of additional work, some parsing and gathering of data from my other sources, combined into a single data-file, I will be able to do this for every country as well. 

The case data and testing data is already available from other sources, but the next step will be to do some of the unique calculations I sometimes do, like my estimates of the % infected, and % currently infected, etc. And that's where the real potential lies. At that point, I may want to think about hosting this somewhere where people can see updated estimates for all these quantities for whatever location they are in. 

COVID-19 Daily Update for Yesterday, Sunday, 8/23/2020:

World Wide Trends in Brief: 

In today's installment of "is it signal or is it noise", let's look at the World Wide daily change in the number of active cases. 

Active Cases are defined as the number of cases - number recovered. So the daily change in active cases, is the number of new cases that day, minus the number of people who have recovered that day. In other words, it's the orange line, minus the green line below:

When it's positive (when the orange line is above the green line) the number of active cases world wide are growing. When it's negative (when the green line is above the orange line) the number of active cases world wide is shrinking. 

When we plot that difference, we see that with a 7 day trend, the daily change in the number of active cases shrank for a while, but is now starting to grow again: 

However, with a 14 day average, the daily change in the number of active cases is still smoothly falling:

Remember that even if this line is heading downward, as long as it is positive, things are getting worse. It needs to be negative before things are getting better. That being said, are things getting worse at a decreasing rate (14 day average)? Or... are things starting to get worse at an accelerating rate (7 day average)? 

This question (is it signal, or is it noise) is one of the KEY questions of statistics. And it is not at all easy to answer. 

Given that there are strong "weekend" effects, we know we should be taking moving averages of the data in increments of 7. (With less than that you start to see weekly fluctuations that make the interpretation even more difficult). However, a 7 day average is potentially too small of a chunk, and it can be impacted by noise which makes you think you are seeing a new "trend" when you aren't. However, the 14 day average is too large, because it can hide trends that are, in fact, real, and you have to wait at least two weeks to be sure that something real has changed. Neither is ideal... and I would prefer something like a 10 day average. But the weekend effects make that even more problematic than either the 7 or the 14 day averages. 

This is a BAD idea, but I will show you what it looks like anyway: 

It is POSSIBLE that the right thing to do would be to create a complicated model of the weekend effect, then take it out of the data, correcting for it... then take a 10 day average, and see what the trend says. But the strength of the weekend effect depends on which country is reporting the most cases at the time (some have a stronger weekend effect than others) so correcting for this is a moving target! It's potentially impossible to really get right. 

So instead... we are just going to have to wait... watch... and see where this trend goes in the next week or so. 

Did I mention that I hate waiting?

via GIPHY

COVID-19 Daily Update for Yesterday, Thursday, 8/20/2020:

World Wide Trends in Brief:

Major world wide trends from my last update are largely continuing. 

Percent daily growth in active cases has been falling, but is now holding at around 0.25%/day:

There have been two days with large outlier reporting spikes that make interpreting the daily deaths difficult, but the trend seems to be down since late July:

Compilation of Data Sources: 

I am in the process of compiling some of the best computer readable data sources for COVID-19 modeling out there. This is a work in progress, but for those interested, here's a link to the document which will likely grow through time. Eventually I hope to wrote code to auto-scrape many of these resources, and I will provide a link to that code once it's available. 

If you are aware of a good data source that I have missed, leave it in the comments here, or feel free to send me an email, or yell at me on FB or Twitter. 

Mis: 

I apologize that today's update is short, I have both a paper deadline, and a deadline for a presentation both due tonight. 

Data: https://docs.google.com/spreadsheets/d/1qVOdkuQ1IQb8McLNoe3oiSrnU1gVj7X916dvpyZ-zZY/edit#gid=0 

COVID-19 Daily Update for Yesterday, Wednesday, 8/19/2020:

Yesterday's update was a "model talk" discussion about problems with rt.live's approach to estimating Rt.

Today is a more traditional status update. 

World Wide Trends in Brief: 

World wide we are starting to see minor signs of improvement. 

Daily cases peaked at the end of July, and have been falling VERY slowly since then. Because recoveries lag, they have continued to climb over the same period:

However, the number of new infections is still larger than the number of new recoveries, which means that the change in the number of active cases is still positive, meaning that more people are sick today than yesterday. But that trend has been moving down over time:

For the world to recover, this trend needs to be negative, and then stay negative. But there are still signs of improvement. 

We can also see improvement in the daily deaths, which have also been trending down since the end of July: 

Can it be Done, Watching China, New Zealand, and South Korea:

We have been especially watching three countries: China, South Korea, and New Zealand to see if their strategies for dealing with the virus is viable. China used a totalitarian lockdown, South Korea used extensive contact tracing with only minor restrictions, and New Zealand "went hard and went early" (without China's totalitarian bent) to stamp out all local transmission before reopening completely. 

All three were initially successful. All three are now facing local outbreaks that are challenging their strategies. If we wanted to prove to the rest of the world that there were alternatives to our failed approaches, these are some important places to watch. 

China's approach of local lockdowns of the affected regions appears to be working: 

New Zealand's second lockdown with contact tracing also now appears to be producing improvements:  

South Korea is still struggling to get its most recent Fundamentalist Church Based outbreak back under control. But they did it once. Now we need to see if they can do it again. It's currently too soon to tell: 

Source, Data, and Graphs:

https://docs.google.com/spreadsheets/d/1qVOdkuQ1IQb8McLNoe3oiSrnU1gVj7X916dvpyZ-zZY/edit#gid=737858658

COVID 19 MODEL TALK, RT.LIVE IS WRONG:

I've been doing daily COVID-19 updates on facebook for a while now, and decided to move them to my blog. 

Today is a statistics and modeling thing, so if that is not your "thing" I promise to do a "status update" tomorrow. And I'll start putting in the title which I am doing each day. 

So... rt.live is wrong: 

https://rt.live runs their algorithm on individual states, but never gives an average for the entire USA. It makes sense that they might not choose to do this, the outbreak in the US is quite diverse, and differs from place to place. But I think it's still a good idea to get the average over-all picture. 

Now that I have their code running on my local machine, I modified the code to run on the average for the US. The first two figures show my results. Remember, this is rt.live's algorithm, but run on data they don't normally show. 

The current estimate of Rt for the US is 0.96, with 80% intervals of 0.73 - 1.13. That seems reasonable. 

HOWEVER, their adjustment for testing rates is VERY aggressive. They estimate that cases were more than TWICE as high in April's peak as they were in July's second peak. 

I don't believe this is right. And if they get the case adjustment wrong, they will get their estimate of Rt wrong in general. I suspect that their CURRENT Rt is right, but their RT back in July is FAR too low. 

If you ONLY look at the daily death curve, you might suspect something like what rt.live is saying. BUT, if you look at the hospitalization curve, it is obvious that this isn't right. 

Instead... the case fatality rate has to be dropping. And that means that the second peak has to be at least as high as the first. I used to claim that the first was larger, because more people were being turned away at the hospitals, but after a conversation with  Youyang Gu over on twitter, I have changed my mind... a falling average age of infection means a lower fatality rate, but ALSO a lower hospitalization rate. And THAT would imply that the second peak should be LARGER than the first! 

Rt.live's second peak is not only not as large or larger than the first, it's MUCH smaller! This just CAN'T be correct. And that means that EVERYTHING else they are doing is also wrong. 

The last figure shows the range of ways that the case data (in blue) can be adjusted for tests. The red curve is Youyang Gu's adjustment (with the second peak much higher than the first). The Yellow curve is my adjustment, with the two peaks roughly equal in size based on the hospitalization data. (I now believe that Youyang Gu's estimate is likely better than mine). The green curve is from rt.live. Not only is it an outlier... it makes NO sense. 

Conclusion: rt.live is wrong. The way they adjust cases for testing rates seems to be FAR too aggressive. For their curve to be right, the Infection Hospitalization Rate (IHR) would have to be RISING DRAMATICALLY, while the Infection Fatality Rate would have to be staying the same. 

The reality is that both the IFR and IHR should be falling as the median age of infection falls, while the IFR should be falling faster than the IHR as treatments improve. 

To get a good and reliable estimate of Rt, I'm going to need to re-write the part of their code that adjusts for testing rates. 

(Here's a link to Youyang Ug's discussion of how he adjusts for testing rates: https://covid19-projections.com/estimating-true-infections/)

Ferguson, Uncertainty, and a Way to Move Forward

I have been silent about the Ferguson matter so far, but I think it's time that I try to articulate a few of my thoughts on this matter.

Before I begin, I want to get something out of the way first. No matter what happened between Darren Wilson and Michael Brown, there is no doubt in my mind that race is an issue in our law enforcement. The reactions of people across the country to this event very clearly demonstrates that fact. People of color in our society simply do not feel protected by the police force that surrounds them, rather, they feel threatened by them. And this is a situation that simply must change moving forward if we want to create an ethical, equitable, prosperous, and peaceful society moving forward. And that fact will remain true, regardless of what actually happened in this single instance between Darren Wilson and Michael Brown.

With regard to those specifics, many people have taken to the internet to tell us exactly what did happen that night, and why they think they know what "really happened". But that is not what I will do. The simple fact is that if Darren Wilson's account of events that night is accurate, then the right decision was reached, and he was innocent of any serious wrongdoing. However, if some of the other eyewitnesses accounts of the events of that night are accurate, then Darren Wilson murdered Michael Brown in cold blood, and a very serious miscarriage of justice has taken place in this instance. And there are extraordinarily compelling reasons not to believe either Darren Wilson's account, or that of the other eyewitnesses testimony. Several of the eyewitnesses testimonies were later refuted by the forensic evidence (for example, testimonies about Michael Brown having been "shot in the back" simply do not match the forensic evidence). The simple fact is that the testimony of witnesses is by far the least reliable source of evidence imaginable. Many innocent people have been sent to prison based upon eyewitness testimony, only to later be exonerated by evidence such as DNA evidence, that simply does not make the sorts of mistakes that eyewitnesses do. And that fact means that both the testimony of Darren Wilson, and that of those others who saw the event are ultimately unreliable.

The result of this, is that I simply do not know if Darren Wilson murdered Michael Brown or not. And I believe that the certainty with which some others (on both sides) have approached this situation is largely unwarranted. So, what am I here to tell you? If I am not here to tell you who to believe, who is right, or whether justice was done, then what am I here to say? I am here to say that I don't know who is right, or what happened, but I do know how to be absolutely sure that this uncertainty does not happen again. I am here to tell you how we can know what happened next time, and how we can make sure that a repeat of this never happens again. And that answer is surprisingly simple.

Every police officer should be required to wear a body camera while on duty and while interacting with the public. Every time. Every police officer. Everywhere. Always. And when this happens we will never again be forced to say that we don't know for sure what happened. We won't have to say that we don't know whether or not justice was done or not when a police officer is not charged in a shooting death. When police officers are innocent of wrongdoing, that will be demonstrated by the camera. When they are guilty of wrongdoing, that too will be shown by the camera. The camera protects both the officer from false accusations, and the public from police abuse. And while knowing what happened after the fact is important, it is perhaps even more important that cameras can actually prevent incidents from ever happening in the first place. Both instances of abuse from police and of bad behavior from those they interact with will go down because both parties will know that they are being recorded, and that the truth of what they are doing will be known. People simply behave differently when they know that they are being watched. And evidence suggests that the use of police cameras can drastically reduce both the incidents of police use of force (up to 50%), and can drastically reduce the incidents of complaints against officers.

Now, this will not solve all our problems with police abuse in this country. And it certainly won't solve all our problems with race in this country either. But it is a start. And we simply must begin somewhere.

This is an idea, who's time has come. Let's make it happen.