The Numbers Game

Posted by

Started out my day taking Rocco to the cancer specialist. The good news is that his blood cancer is very well controlled and the oncologist has recommended decreasing his medication again. The weird thing was arriving in the parking lot and being met by a vet tech in full PPE (masks and gloves). Then having to hand Rocco over to her because I couldn’t enter the premises.

Now, this happened when we went to the regular vet the other day but the tech there didn’t have a mask and gloves. Normally, you’d imagine that they were trying to contain some plague within the building but this was to keep the plague out. I don’t know why it struck me but it was all because of me and not Rocco. I’m more of the threat. I know that’s obvious in the world of today but it was just one of those little things that hit me.

I’ve previously mentioned that I do have a degree in Statistics and I have free time and numbers so I’ve again been playing with them. I finally saw another website that had their own set of projections which painted a pretty grim picture for the next weeks. Now I’m just a guy playing with Excel and they are probably using much more sophisticated techniques with better knowledge about the spread of plagues, but things just don’t add up.

I think they had a figure of 95000 deaths. According to the CDC, we’ve had 5443 as of today. Italy is absolutely the hardest hit country in the world and they are at about 13000 deaths as of yesterday. I mean it is certainly possible that things could suddenly sky rocket. I know you’ve got the 14 day incubation and asymptomatic carriers spreading it but it just doesn’t add up.

The death rate is going up. Some of that is likely due to over taxed hospitals and insufficient equipment. I’ve heard and I think it is confirmed that some of this is because they are pretty much counting anyone who tested positive for the virus as dying from the virus even if there were plenty of other co-morbidities. I’ve seen the phrase “complications from the virus” which seems like kind of a catch all. Anyway, even with that, the calculated fatality rate is about 2%.

For the sake of argument, lets just assume that goes up to 5%. To get to 95000 deaths, you’d need 1,900,000 people infected. Is that possible? Yes, if the cases continue to increase at the current rate, then we’d hit that figure some time in early May. Is that possible? Of course, but it assumes we’ll see no slow down in the rate of new cases for a month.

The problem for me in projecting this is that we don’t yet have a lot of experience with what happens after the rapid increase. For a whole lot of reasons, I don’t trust the numbers out of China and that really leaves South Korea as the only place where you have (a) a significant number of cases and (b) clear evidence that number of new cases has dropped. Extrapolating from one place isn’t a smart idea.

There are hints of potential good news in the data. Italy has clearly been the hardest hit country but the number of new cases there has been dropping since March 27th. As a caveat, this is based on WHO statistics and I don’t have any idea what’s going on in Italy with respect to testing so there could be all kinds of problems with the numbers. It is certainly true that Spain is going to pass them for total number of cases, so there does seem to be evidence of a slow down.

But a slow down is different than getting to a manageable number like South Korea. (Assuming their 100 new cases a day is “manageable”). Italy has been at a peak for about two weeks and we have to see how long it takes to drop to know what the potential course might be here.

I did run a model based on the observed data on Tuesday. It has predicted the number of new cases to within .5% for the rest of this week. Again, unless the CDC is hiding some secret numbers, it still doesn’t suggest that things are suddenly going to start going towards the moon. But, now that I have a model and projections, I can continue to test it as new data arrive.

And because I have a lot of free time, I did this for all the states. Again, this is based on CDC data (which they get from the states). At the state level, there are holes in the data and probably more noise around the actual figures. Still, I got some decent models. I used them to predict new cases for today.

Using a 5% error as a “good” prediction, there were 13 states where the model didn’t fit. Not great but not wildly off either. Four states (Alaska, Montana, Vermont and Connecticut) had fewer cases than their models would have predicted. But nine states (Pennsylvania, Tennessee, Virginia, Oklahoma, South Carolina, Nebraska, South Dakota, Idaho and Louisiana) had more than would have been predicted. Louisiana was the worst. They saw a huge jump from yesterday to today.

What does that mean? Well, the number of cases in the US is still dominated by New York. This suggests that we are seeing the rate of cases pick up in other places and that could send the overall numbers higher than I’d predict. On the other hand, we are still talking about a small number of cases for many of these states so they’d have to really crank things up to make a difference in the national total.

Speaking of New York, the other things I’ve looked it is the rate of change in the number of new cases over the week. For the US as whole, we went from 140907 on Monday to 239279 today. These are the CDC numbers and they are actually one day behind but I’m using them to be consistent. That’s an increase of 70%. In New York, they went from 59219 cases on Monday to 90279 today. That’s an increase of 52%. Not great but behind the national average. Which may mean New York is getting close to a peak while other places are ramping up. Just something else to monitor.

I’ve taken a different set of numbers and tried to do some projections for other countries as well. Right now, most of the world is like the US and in a state where the cases are still rapidly going up. But I’ve projected out about a week and then I can see if these model continue to hold up. If there is anything interesting that comes from it, I’ll let you know.

Sorry, for all the stat geek stuff. As I’ve said, I’ve got lots of time on my hands and I’ve always liked playing with numbers.

I will say that after doing all this, I’m still convinced we’re going to get through this. But it is likely going to take longer than any of us want. I don’t know where the final US numbers are going to end up but it would not shock me at all if we have over a million cases by the time all is said and done.

I understand if you feel this way!

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.