Sunday, May 31, 2020
COVID-19 May 31, 2020
There is nothing remarkable on this last day of May. If I try really hard, I might find a hint of death rates failing to fall as fast as the model suggests that they should. I'm not going to make this call just yet, but it's something I'm watching. Since the number of new cases has been falling very slowly, I'd expect the deaths to mirror that with a lag. Perhaps that's starting to happen. Perhaps. I'm also seeing an increasing number of states setting records for the largest number of new cases. I'm not going to get too excited about that either. Not yet. As testing increases, we might expect a larger number of confirmed cases. On the other hand, it probably also means that the spread of the virus isn't exactly waning. So, there's a couple of things to watch.
to watch over the next week or two.
Saturday, May 30, 2020
COVID-19 May 30, 2020
No surprises today. Everything continues to follow established trends. That is all.
Friday, May 29, 2020
COVID-19 May 29, 2020
TGIFF. :) Numbers from yesterday are right in line with the longer-term trends. The death rate is slowly decreasing, but we're still losing many hundreds of our neighbors every day to the virus. Daily infections are decreasing on average as well even while a few states seem to have leveled off or may, possibly, be increasing.
My data source still shows us under 100,000 total deaths. I may have found some reasons why this data set is low compared to several others. First, because it relies on the state data, it does take into account revisions to the data that could be days or weeks old. For example, in my State of Colorado, there's a note that says, " On April 25, 2020, Colorado removed 29 deaths. All 29 removed deaths were probable deaths which were entered as duplicates." Were these data removed from the other data sources or is their data locked in once it is entered? Colorado is not the only state that has these corrections. I like my data set because all of these modifications are fully documented. Secondly, and this may be the most important, some of the data sources are counting federal prisons as a separate category from the state data. However, I have a feeling these data are being double-counted. This is something that requires additional research. It may also be that my data is missing these data points. Either way, that could explain some of the difference. Finally, and importantly, some states only count confirmed COVID deaths while other count confirmed and presumptive COVID deaths. My data set follows whatever the states do. Other data sets do not or may not.
My data source still shows us under 100,000 total deaths. I may have found some reasons why this data set is low compared to several others. First, because it relies on the state data, it does take into account revisions to the data that could be days or weeks old. For example, in my State of Colorado, there's a note that says, " On April 25, 2020, Colorado removed 29 deaths. All 29 removed deaths were probable deaths which were entered as duplicates." Were these data removed from the other data sources or is their data locked in once it is entered? Colorado is not the only state that has these corrections. I like my data set because all of these modifications are fully documented. Secondly, and this may be the most important, some of the data sources are counting federal prisons as a separate category from the state data. However, I have a feeling these data are being double-counted. This is something that requires additional research. It may also be that my data is missing these data points. Either way, that could explain some of the difference. Finally, and importantly, some states only count confirmed COVID deaths while other count confirmed and presumptive COVID deaths. My data set follows whatever the states do. Other data sets do not or may not.
Thursday, May 28, 2020
COVID-19 May 28, 2020
Last week it appeared that the 7-day average of the new case rate had started to slowly increase. That was, apparently, just a little bump. This week the 7-day average has leveled off again and even decreased a bit more. As of today, the 7-day average has dropped to 20,852 and the best fit 7-day line has dropped below 20,000 cases/day for the first time in a long time. (The equation of the best-fit line is shown in the cumulative case plot. The slope of the line is equivalent to the case rate and is 19,945. Algebra. Again.). These trends are despite an increase in testing and loosening of restrictions. Will it last? I'm not a betting man, but if I was, I know where I'd put my money.
The death rates continue their downward trend, at least when looking at the 7-day average. The usual peaks and valleys in the daily numbers are still there. Yesterday was an "up" after three days of "down".
Wednesday, May 27, 2020
COVID-19 May 27, 2020
Really busy today. Nothing really new to report. Things are more or less the same as yesterday, but more of it.
Tuesday, May 26, 2020
COVID-19 May 26, 2020
I would have thought that we'd be seeing a noticeable upward spike in new cases and even deaths by now. But, it hasn't happened. Let's hope it never comes. I wouldn't bet the farm on it though.
Daily deaths continue a nice average decrease and are following the Gompertz model. I didn't expect such good agreement this late into the pandemic. Per yesterday's post, I note that the data I'm using shows us at ~92,000 deaths which is many thousand away from the grim 100,000 mark. Your data source may be different. I have no way of knowing what data set is correct. They can't all be correct and yet different at the same time. They're probably all wrong. What I can say is that my data is fully traceable and verifiable to the source. Based on the Gompertz model and using my data I show 100,000 on or about June 8. Rest assured the media will find the most grim data set they can and use that to claim we've hit 100,000. What they won't do is pay much attention to the reliability of that data. Nevertheless, 100,000 we will hit soon.
Speaking of media, I continue to see reports about how some state had a big daily increase in the number of cases. That may be technically true, but given the noise in the day-to-day statistics such daily data are little more than curiosities. It's the longer trend that matters. If high numbers continue, on average, it means something. Beware of media porn.
Here are today's stats and plots.
Daily deaths continue a nice average decrease and are following the Gompertz model. I didn't expect such good agreement this late into the pandemic. Per yesterday's post, I note that the data I'm using shows us at ~92,000 deaths which is many thousand away from the grim 100,000 mark. Your data source may be different. I have no way of knowing what data set is correct. They can't all be correct and yet different at the same time. They're probably all wrong. What I can say is that my data is fully traceable and verifiable to the source. Based on the Gompertz model and using my data I show 100,000 on or about June 8. Rest assured the media will find the most grim data set they can and use that to claim we've hit 100,000. What they won't do is pay much attention to the reliability of that data. Nevertheless, 100,000 we will hit soon.
Speaking of media, I continue to see reports about how some state had a big daily increase in the number of cases. That may be technically true, but given the noise in the day-to-day statistics such daily data are little more than curiosities. It's the longer trend that matters. If high numbers continue, on average, it means something. Beware of media porn.
Here are today's stats and plots.
Sunday, May 24, 2020
COVID-19 May 25, 2020
Let's talk about data! Not the data from yesterday, but the origin of the data. A lot of you posted on my Facebook site questioning the veracity of the death data. Several others contacted me privately. Let me say, "Thank you!" Data is king and it is absolutely proper to question the data I'm providing. I have been as transparent as possible about the source of my data from my very first post. I even include the link on the plots I put out daily. Here it is again: https://en.wikipedia.org/wiki/Template:COVID-19_pandemic_data/United_States_medical_cases
"Wait! You're using Wikipedia as a data source? That's not a reliable source!". You're right. It isn't. At least by itself. I'll come back to this at the end. But first, let's look at some other possible data sources: The CDC, Johns Hopkins University, Worldometer. I'll say a little about each of these and why I have rejected them as a source in favor of Wikipedia.
The CDC (U.S. Center for Disease Control): The CDC Covid-19 website is here. If you poke around a little bit, you'll find a page with a graphic that looks like this:
John Hopkins University info is here. It's a really fancy interface, albeit sometimes painfully slow even on my fairly beefy desktop. There's a section on that site that shows total deaths. The screenshot is below:
It shows a number closer, but still different than the CDC. Perhaps they've incorporated numbers from today whereas the CDC has not yet done so? Who knows. And I mean that literally. Who knows! I don't, and neither do you, because try as I might, I was unable to find any statement of where their data is coming from. That's amazingly sloppy work for such a respectable institution. Presumably, they aren't collecting it themselves, so like the CDC I assume they are a data aggregator. Are they getting it from the same 55 jurisdictions as the CDC? Are they getting it directly from the CDC? Who knows. I've rejected this site as a source of data because it provides no traceable way to verify the data. I don't know where it came from nor can I verify that the data has been tabulated properly.
How about Worldometer? They are currently showing 99,300 deaths. Different and higher again. At least they have some info on the source of their data: "Our sources include Official Websites of Ministries of Health or other Government Institutions and Government authorities' social media accounts. Because national aggregates often lag behind the regional and local health departments' data, part of our work consists in monitoring thousands of daily reports released by local authorities. Our multilingual team also monitors press briefings' live streams throughout the day. Occasionally, we can use a selection of leading and trusted news wires with a proven history of accuracy in communicating the data reported by Governments in live press conferences before it is published on the Official Websites." So, I want to verify the Worldometer, like the CDC, I have to go hunting for the data, but this time they don't even tell me exactly where. I rejected this data for exactly that reason. Like JHU, I don't know where the data came from and I am unable to verify that the data has been tabulated correctly.
Let's loop back to Wikipedia. That site is an aggregator like the others. Unlike the others, however, they have comparatively exquisite detail on all the data. It is broken down by state and territory and further provide on a daily basis. There are direct links to the sources of the data--usually the state health departments. Further, there are now hundreds of footnotes describing any irregularities or idiosyncracies about the data, almost all of which can be traced directly to the source of that information. This is a nice data set, but it doesn't mean it's correct.
I have selected Wikipedia as a data source because it is fully traceable to the origin of the data. All the information is fully documented. And, importantly, every time I do a spot check, I find the data to be exactly as reported by the individual states. For example, this evening I see that New York is reporting 109 new deaths for a total of 23,391. The New York health department shows the same exact values for the day and for the total. I find the same agreement for Texas. And Colorado. And others. Exact. And it's not just this time, but every time I've checked. To the extent I can do this for the CDC data and JHU data, I find that they DO NOT agree with the state data. Why? I don't know, because they don't provide any information on their data sources.
Now, if I take Wikipedia as a good and verifiable data set, I get the numbers I display every day. For today that was 87,372. As far as I can tell that matches the data provided by the states. Yes, it's lower than many other sources and is lower than what's been touted in the press (Note: the press is not a reliable source of data). But, as the CDC says, you should trust the states' numbers over the CDC's numbers.
Why such a difference? I don't know for sure, because the CDC doesn't provide numbers in a way that I can easily check against the states' data. I do have some ideas. It is not uncommon for the states to adjust their data days and sometimes a week or more after the numbers are reported. I adjust my numbers accordingly. Sometimes states have double counted deaths for example. Perhaps the CDC took those numbers, locked them in, and then never went back and fixed them once they were revised. It seems unlikely that alone could explain the large discrepancy between the numbers I'm using and the CDC. Your guess is as good as mine. We'd know if all the institutions practiced good and transparent data management. But they don't.
One of the tenets of science is reproducibility. Anyone should be able to conduct your experiment or crunch the same numbers and get the same answer. The only site that allows this is Wikipedia. So, is my data correct? I don't know. I do know that it is reproducible. I know that anyone can check my data and go straight to the source to verify the values.
Having said all that, here's the data for today. It came in while I was typing this up so I'll post it now instead of waiting until tomorrow.
"Wait! You're using Wikipedia as a data source? That's not a reliable source!". You're right. It isn't. At least by itself. I'll come back to this at the end. But first, let's look at some other possible data sources: The CDC, Johns Hopkins University, Worldometer. I'll say a little about each of these and why I have rejected them as a source in favor of Wikipedia.
The CDC (U.S. Center for Disease Control): The CDC Covid-19 website is here. If you poke around a little bit, you'll find a page with a graphic that looks like this:
If you compare these numbers to what I had for May 24 (which are actually numbers accumulated as of May 23) I have 1,61,2018 for total cases and 87,372 for total deaths. Neither of my numbers agree with the CDC, but the deaths appear to be substantially different, as many of you point out. If you click on the little "About the Data" link below the Total Deaths statistic on the CDC site, you can learn a little bit more about the origin of their data. There are a couple of important things. First, it says, "Numbers reported on Saturdays and Sundays are preliminary and not yet confirmed by state and territorial health departments. These numbers may be modified when numbers are updated on Mondays." I'll point out that today is Sunday. That means yesterday was Saturday. Second, and more importantly, it says, "State and local public health departments are now testing and publicly reporting their cases. In the event of a discrepancy between CDC cases and cases reported by state and local public health officials, data reported by states should be considered the most up to date." So, right there on their own site they tell you that the data from the states should take precedence. Finally, the CDC says, "There are currently 55 U.S.-affiliated jurisdictions reporting cases of COVID-19. This includes 50 states, District of Columbia, Guam, the Northern Mariana Islands, Puerto Rico, and the U.S Virgin Islands." So, at least we know where they claim to get their data from. They even have a map that you can click on that takes you to each of the 55 jurisdictions. Basically, the CDC is a data aggregator and it's up to the user to go and check all the individual sources of data and make it's consistent. I don't have the time for that, and further, the CDC doesn't make it easy to see all the daily data from all the 55 jurisdictions. As I'll show when I get back to Wikipedia, I have reason to believe the CDC has not tallied things correctly. Nevertheless, if one takes the CDC data without digging further to the actual source, you get the numbers above.
John Hopkins University info is here. It's a really fancy interface, albeit sometimes painfully slow even on my fairly beefy desktop. There's a section on that site that shows total deaths. The screenshot is below:
It shows a number closer, but still different than the CDC. Perhaps they've incorporated numbers from today whereas the CDC has not yet done so? Who knows. And I mean that literally. Who knows! I don't, and neither do you, because try as I might, I was unable to find any statement of where their data is coming from. That's amazingly sloppy work for such a respectable institution. Presumably, they aren't collecting it themselves, so like the CDC I assume they are a data aggregator. Are they getting it from the same 55 jurisdictions as the CDC? Are they getting it directly from the CDC? Who knows. I've rejected this site as a source of data because it provides no traceable way to verify the data. I don't know where it came from nor can I verify that the data has been tabulated properly.
How about Worldometer? They are currently showing 99,300 deaths. Different and higher again. At least they have some info on the source of their data: "Our sources include Official Websites of Ministries of Health or other Government Institutions and Government authorities' social media accounts. Because national aggregates often lag behind the regional and local health departments' data, part of our work consists in monitoring thousands of daily reports released by local authorities. Our multilingual team also monitors press briefings' live streams throughout the day. Occasionally, we can use a selection of leading and trusted news wires with a proven history of accuracy in communicating the data reported by Governments in live press conferences before it is published on the Official Websites." So, I want to verify the Worldometer, like the CDC, I have to go hunting for the data, but this time they don't even tell me exactly where. I rejected this data for exactly that reason. Like JHU, I don't know where the data came from and I am unable to verify that the data has been tabulated correctly.
Let's loop back to Wikipedia. That site is an aggregator like the others. Unlike the others, however, they have comparatively exquisite detail on all the data. It is broken down by state and territory and further provide on a daily basis. There are direct links to the sources of the data--usually the state health departments. Further, there are now hundreds of footnotes describing any irregularities or idiosyncracies about the data, almost all of which can be traced directly to the source of that information. This is a nice data set, but it doesn't mean it's correct.
I have selected Wikipedia as a data source because it is fully traceable to the origin of the data. All the information is fully documented. And, importantly, every time I do a spot check, I find the data to be exactly as reported by the individual states. For example, this evening I see that New York is reporting 109 new deaths for a total of 23,391. The New York health department shows the same exact values for the day and for the total. I find the same agreement for Texas. And Colorado. And others. Exact. And it's not just this time, but every time I've checked. To the extent I can do this for the CDC data and JHU data, I find that they DO NOT agree with the state data. Why? I don't know, because they don't provide any information on their data sources.
Now, if I take Wikipedia as a good and verifiable data set, I get the numbers I display every day. For today that was 87,372. As far as I can tell that matches the data provided by the states. Yes, it's lower than many other sources and is lower than what's been touted in the press (Note: the press is not a reliable source of data). But, as the CDC says, you should trust the states' numbers over the CDC's numbers.
Why such a difference? I don't know for sure, because the CDC doesn't provide numbers in a way that I can easily check against the states' data. I do have some ideas. It is not uncommon for the states to adjust their data days and sometimes a week or more after the numbers are reported. I adjust my numbers accordingly. Sometimes states have double counted deaths for example. Perhaps the CDC took those numbers, locked them in, and then never went back and fixed them once they were revised. It seems unlikely that alone could explain the large discrepancy between the numbers I'm using and the CDC. Your guess is as good as mine. We'd know if all the institutions practiced good and transparent data management. But they don't.
One of the tenets of science is reproducibility. Anyone should be able to conduct your experiment or crunch the same numbers and get the same answer. The only site that allows this is Wikipedia. So, is my data correct? I don't know. I do know that it is reproducible. I know that anyone can check my data and go straight to the source to verify the values.
Having said all that, here's the data for today. It came in while I was typing this up so I'll post it now instead of waiting until tomorrow.
COVID-19 May 24, 2020
The pandemic marches on. New cases continue to hold steady or rise slightly, but the death rate continues to fall and continues to follow the Gompertz model remarkably well. If that agreement continues, we will hit 100,000 deaths on or about June 8. On that date the model predicts 444 deaths per day, which is less than half the current rate.
Saturday, May 23, 2020
COVID-19 May 23, 2020
No significant changes in the trends to report. Here are the data. Go enjoy the weekend!
Friday, May 22, 2020
COVID-19 May 22, 2020
How about some not so bad news to kick-off the holiday weekend? The daily death statistics are still on the decrease. When I saw the minimum values earlier this week I was concerned that they hadn't dropped enough to keep the downward trend going. But, this week's peak was also lower than I would have expected, so on average things are still going in a good direction. I adjusted the Gompertz model very slightly upward from 100,000 to 108,000 total deaths in order to better match the cumulative death data. This assumes that everything continues as it has in the past. I doubt that it will, but for now we are still following the model.
In contrast, new cases are now clearly showing the average upward trend. I don't have anything more to say about that.
Thursday, May 21, 2020
COVID-19 May 21, 2020
I have another very busy day today, but I didn't want to skip another day. This is going to be short. There's no significant change in the trends. The death rate has flattened is essentially flat. New cases have started to move up very slightly. I'm waiting for the weekly peak, which should occur today or tomorrow.
Tuesday, May 19, 2020
COVID-19 May 19, 2020
We've passed through the minimum for the week, and the trends are starting to look more worrisome. Let's start with the Daily New Cases. Prior to the last week, the maximums and minimums had slowly but steadily been decreasing. As a result, the number of new cases decreased, on average from about 30,000 per day on April 23 to about 23,000 per day on May 14. I had noted that this past week's maximum was about the same as the prior week and was skeptical that the upcoming minimum would be low enough to compensate. Well, the results are in, and not only was the minimum not low enough, but it was higher than last week's minimum. The trend is now broken and the 7-day average is back on the uptick. Higher numbers are very likely to come in for the rest of the week if they follow the weekly cycle. This will drive the 7-day trend even higher. I suspect this is just the beginning of what's to come.
Turning to the Daily Death statistics, it looks like the 7-day trend is heading toward a more steady number and that the slow decrease is also approaching an end, but that signal is less robust than for the Daily New Cases. Since the number of deaths should be a function of the total number of cases, it makes sense that the death statistics should start to reflect the infection statistics. Unfortunately, the relationship between these two data isn't as tight as it should be, because the availability of testing has an impact on the number of cases that are reported.
Turning to the Daily Death statistics, it looks like the 7-day trend is heading toward a more steady number and that the slow decrease is also approaching an end, but that signal is less robust than for the Daily New Cases. Since the number of deaths should be a function of the total number of cases, it makes sense that the death statistics should start to reflect the infection statistics. Unfortunately, the relationship between these two data isn't as tight as it should be, because the availability of testing has an impact on the number of cases that are reported.
Subscribe to:
Posts (Atom)