Publisher’s Note: Yesterday (March 23) we began posting a series of charts by Douglas G. Frank, a “scientist, inventor, entrepreneur, libertarian, teacher, musician, daddy, husband, and servant of Christ.” He is also President of Precision Analytical Instruments, Inc., which “specializes in the design, manufacture, and development of custom analytical and control instrumentation.” We continue that series today, as we will each day until the model proves unsustainable or the crises diminishes. Mr. Frank’s work is part of a long-line of thinkers who look at data, analyze the data into statistics, and use the results to make informed decisions. Our prayer is that it provides comfort and insight.
All words below this line are compiled from Mr. Frank. Other than ordering and organizing the comments and the charts, we (DPH) have not edited any of his comments because that would be out of our area of expertise. The comments were given in a social media context, so have a friendly, conversational tone.
Covid-19 “A Note About Data Smoothing”
I have avoided “smoothing” any data so far, for fear of “disguising” what is really going on. To scientists, data are sacrosanct, and any fiddling is forbidden without careful disclosure and proper statistical criteria being met.
For example, the massive swings in the California data are *real.* They reflect massive swings in the testing protocols of California. If I’d smoothed the data before I showed it to you, then I would have ended up “hiding” what is actually going on.
So especially at the start, I prefer to present raw data. But the algorithm uses a smoothing algorithm in the background to do the analysis, because derivatives (sorry, calculus terms) really, really hate sharp turns in curves, and it messes up the analysis. Well, once I fixed that glitch, then lo and behold… POW, the California data are amazing.
Ok, so now and then I am going to show you smoothed data in my graphs… but I will always tell you, so you know.
From Doug Frank on State Shutdowns:
Every parent understands this situation:
1) Your child is playing in the street. You rush out to remove them from the street and scold them.
They say, “But there were no cars, it was safe.”
You say, “Maybe that was true *this time,* but it might not be true next time. Stay out of the street!”
They conclude: (“I’ll be more careful next time.”)
2) Your child is playing in the street. You rush out to remove them from the street and scold them. “I told you not to play in the street! It’s dangerous!!!”)
You repeat the same litany with them, and they come to the same conclusion.
3) This happens again, and again, and again.
Now, you are faced with a dilemma. You know they are going back into the street, that a car could hit them, and they would be dead. So you either have to build a fence or keep them locked in the house. The risk is just too great.
Welcome to state shutdowns.
World and USA Model:
[My wife] was telling me tonight that the WHO was hyping the fact that the total world numbers are suddenly accelerating, and that we all need to do more!
Don’t get me started. Students! How many statistical fallacies can you find in this statement!
Frankly, it is appalling.
I added today’s data to the world graph. Looks like some adjusting will be required. But it was only my first stab at it.
Covid-19 “Big US Model Update”
After updating the states with the highest numbers of cases, here is the “Big US Model.” Essentially, it combines all the state stats into one big peak, so we know how the country is doing as a whole.
Because of the New York situation, the peak has slipped a day later, to 3/29. Most of the states will have just peaked or be peaking then. Hopely those days don’t seem so scary to you anymore.
The forecast for tomorrow is 13,200 new cases. My gut is telling me this number is high, since we didn’t reach the projection and since so many states seem to be flattening.
Covid-19 “Quick Look at the US”
Just added the new data, no adjustments. Any doubt we are tracking a peak?!
Covid-19 “US Deaths Tracking”
Covid-19 “US Total Forecast, Recalibrate”
Folks, several of you are asking me about it, so I need to address it. (“We interrupt this program to bring you a special news bulletin… lol)
**I like this** because it means you understand what I am doing. If you’ve been tracking me, you know that I will always tell you the truth. And when I’m wrong, I’ll admit it, and tell you exactly what I am going to do about it.
My original forecast for the whole country was about 500-1500 total deaths. Clearly that was optimistic, probably by a factor of two or three. That “back of the envelope” forecast was based upon the testing data that was available at that time. Later, when the testing dramatically went up, the results painted a different story. I mentioned it at the time, and even said that I was reluctant to change anything until the testing protocols settled in and the stats became reasonable again. Well now that the stats are settling in, we need to recalibrate our thinking. But maybe not the way you think.
The scientific method demands that when our hypothesis fails (e.g. 1500 deaths) we must modify our fundamental assumptions.
Right now, it is looking like New York *alone* will exceed my original forecast. That is sad, not because my forecast will be wrong, but because that is a lot of lost lives. Pray for New York.
And what is happening in “New York” is really “NEW YORK CITY.” I have info from boots on the ground there that testing in most of the counties is coming up remarkably clean. **One city** is doing this, not the whole state.
And the vast majority of the rest of the country is still in line with my original intuitions and calculations. Even today we are revising numbers DOWN.
So don’t let the numbers confuse you… a big increase in *total* projected deaths in one city does not mean the whole country is going to flinders. To the contrary, the NY data are closely following a very predictable trend, and so are nearly all the states we’ve modeled so far. For the most part, we still know when things are coming, and how much to expect.
Now what to do about it. I will think on that and get back to you later. There are several possibilities.
All along, I’ve said that the most reliable way to track this would be the “deaths” graph. Here is the one for today. I did not make any adjustments. Just added the data. As I discussed earlier, in light of new developments we have some adjustments to make.
State by State (Alphabetical)
Covid-19 “Quick Look at California”
Waited for the West Coat Updates to update California (I’ll also go back and revise Washington and Oregon if necessary) The only way I could get the California data to reconcile with the model was to bring DOWN the number of cases. Not a bad thing! So, I did it. Could it be? California is reaching it’s peak? If so, this would be fabulous for us, because California would be over the hump and waning while the rest of us are still climbing the peak. A perfect example to follow… over the peak and DOWN.
Fingers crossed. Sometimes noise in the data can confuse the model. But heck, I just turned down the knob on ten other states, why not California too? Also, California was one of the earliest infections. They SHOULD be ahead of the rest of us. But should is not reality. But so far, so good.
****** ALERT *****
Thanks to the sharp eyes of our community, I revisited the data, and discovered that the AI was struggling with that nasty glitch in the cases data. So, I smoothed it the cases data and reapplied the model. Cha-Ching!
Covid-19 “Quick Look at Colorado”
Woo Hoo! Another state with cases adjusting down. So, I had to tweak up the mortality rate a tad. Kind noisy still, but looking good!Keep it up, Colorado!
Covid-19 “Quick Look at Florida”
These stats are so good… makes it easy! Still, it is just my first stab at Florida. Florida’s website is way-cool, but NO graphs. Interesting. Just current numbers. Hmmm….
Florida clearly has a well-established infection. When the Spring-breakers return, watch out for the secondary infections.
Covid-19 “Quick Look at Illinois”
This epidemic is almost entirely Chicago. Nevertheless, another state with LOWER numbers.
So I slightly tweaked the numbers, but things can still change. Hopefully, the wind in this windy city is blowing North…
Covid-19 “Quick Look at Indiana”
Took the cases down 500 (from 3000) and had to raise the mortality rate a tad to compensate. Tracking well.
Covid-19 “Quick Look at Kentucky”
ANOTHER state I’m adjusting down today. Look at that drop in case numbers! These numbers are still noisy… bet they keep going down.
Covid-19 “Quick Look at Louisiana”
Another state, adjusted DOWN! I reduced the cases by 500. Otherwise, tracking like ants following a candy trail. Folks, ain’t it a thang to behold! (And obviously, some kind of glitch in the cases reporting)
Covid-19 “Quick Look at Maryland”
Wow. Fabulous stats for Maryland. Right on the dot. My first crack at it, but with stats fitting so well, the forecast is pretty reliable. I wonder, did they do LOTS of testing in Maryland?
Covid-19 “Quick Look at Massachusetts”
Didn’t change a thing, just inserted the new data. Yet another state with lower numbers (woo hoo!), but notice that the TOTAL number is right on the projection (Fermi to the rescue!)
Covid-19 “Quick Look at Michigan”
Covid-19 “Quick Look at Minnesota”
My first look at Minnesota. Chose some reasonable starting assumptions. Noisy data.
Covid-19 “Quick Look at New Jersey”
Great stats here. Only tweaked the parameters a tad.
Covid-19 “Quick Look at New York”
I’d been dreading this update today, because NY is suffering. NY alone may equal my initial prediction for the entire country. But the silver lining is that the model NAILED it today. I didn’t change a thing, just added the DATA. Each day our forecasts are becoming more reliable, precisely as we anticipated.
Start spreadin’ the news…
Covid-19 “Quick Look at North Carolina”
All I did was adjust the total cases down by 500 (from 3500) and added the data. Ho hum, ANOTHER state with fewer projected cases. Nice! Oh… and since there are ZERO deaths there, the death rate remains an initial assumption.
Covid-19 “Quick Look at Ohio”
Hot new numbers just released for Ohio. Looking good Ohio! I adjusted the total cases down a sniggle, to 2200 from 2500, and so had to increase the mortality up a sniggle to 1.2%. This is the third state I’ve adjusted DOWN today. This is likely due to the rapid increase in testing settling in. At first, it finds lots of cases, but then positives become more rare. Perfect!
Covid-19 “Quick Look at Oregon”
I didn’t change anything… just added the new data. Bullseye… right on the money. Except, I brought DOWN the mortality rate a tad from 3% to 2.5%. All good!
Covid-19 “Quick Look at Pennsylvania”
Look at these beautiful stats. It makes my mouth water. Because the stats are so good, I made a different kind of adjustment to them than usual; I *sharpened* the peak. This means that the peak gets here a little earlier, but the epidemic is over sooner! The FACT that these are tracking so well is reason for celebration!!! Why? Because it means that everything is progressing as it should, and we are more confident when it is all gonna end. Of course, we are very sad for our lost ones, I’m not diminishing that, but sitting in my chair, I am exhuberant!
Covid-19 “Quick Look at S Carolina”
Had to adjust this one up a bit. Not because anything is happening, but because my original stab in the dark was low. The data are still very noisy!
Covid-19 “Quick Look at Tennessee”
Didn’t change a thing, just updated the data.EXCEPT, I had to turn down the mortality rate. Oh, yeah.
Covid-19 “Quick Look at Washington”
Another state I get to adjust DOWN today. I am liking it. The epidemic in Washington is one of the earliest infections, and has been long and drawn out. This is certainly a “flattened” curve. But I don’t think I like it much. I want it over. If I were a political leader, the last thing I would want is a long, drawn-out epidemic. Fast and furious, get ‘er done. Maybe all the state crack-downs is the right idea.
Covid-19 “Quick Look at West Virginia”
West Virginia (new chart): Here is my biggest shot in the dark yet.
Covid-19 “Quick Look at Wisconsin”
Big drop in cases…. following the trend, so I DOWN adjusted. Woo Hoo!
Covid-19 “Quick Look at Australia”
Only added data, didn’t change a thing, except I got to decrease the death rate (Yesssss!). Talk about your testing artifacts! Whoa! They are tracking great. At their worst, and the numbers are small. Say a prayer of thanks for Australia.
Covid-19 “Model Update for S Korea”
Just added the data. Notice that their secondary infection (red squares) is leveling off. Perfect: a single secondary infection based in a elder care facility was contained. Much higher death rate there, of course. We are pleased that S Korea was on top of this. We need to learn from their example. I like having examples like this available, because things like this are certain to happen in the US, and our eyes will be trained to recognize them in the data.
Covid-19 “Italy Model Update”
Every day before I update Italy, I say a prayer. God help these people. 703 people died there today from this. The silver lining is that my initial assumptions about the secondary infection were really close, so the deaths plot is right on the money, so we know where they stand, and there is a light at the end of the tunnel. I’ve included a plot of the secondary infection in the graph, so you can see that their secondary infection essentially spreads out their peak. At precisely the time they were already at their peak. A perfect storm.
Pray for our friends in Italy. God grant them peace.
Covid-19 “Quick Look at India”
Well, so far my sense of “impending doom” India is unwarranted. Thank God! This is my first look at India.If you ask me to explain what I am looking at, I would hypothesize that they had a small initial infection *which they detected and contained.* And have since practiced good hygiene and distancing measures. Kudos appear to be in order.I hope that I am right, and I pray it holds. C’mon, India!
Covid-19 “Model Update for Iran”
Everyone pay attention!!! This is a FABULOUS example of what I have been describing. All I did was add the data.
Note that the “cases” is diverging recently, but the “deaths” is on the money. Clearly, Iran has increased its testing protocols… makes sense; the WHO is asking everyone to do it, and they are in clean up mode… Iran doesn’t want any secondary infections. So the increase in “cases” is not an *actual* increase in cases, merely an increase in the *fraction* of actual cases detected. Shows why the deaths curve is really the right way to track progress.
Hooray for Iran! Keep it up!!!