Testing Testing Testing and How is Florida Doing?

If you watch the news lately (and what else do you have to do?), you probably have heard the next hurdle the country faces is conducting more Covid-19 tests. More testing helps identify people with the virus, including asymptomatic individuals walking around out there who don’t know their contagious. You know, the guy sitting next to you on the bus.

Finding infected people also allows for contact tracing, or finding and testing the people who have been recently in contact with a positive individual.

But how many tests do we need? Some experts say we need a minimum of a 152 tests per 100,000 population per day. In Florida, that is nearly 32,000 tests a day!

Other experts say that the number of positive cases should be at or below 10%. The logic is that if the number of positives are higher 10%, then you are not doing enough testing to capture all the infected people in the community.

Florida residents, up to this point, have done pretty well with social distancing, as the number of new Covid cases have slowly declined, as our tracking models show. But before we really return to normalcy without a vaccine, we need to know who has already had the virus, those who are asymptomatic and those who are still infected.

So let’s first start of how Florida is doing with conducting tests as of today. According to the Covid-19 Tracking Project as of April 24th, Florida has conducted a total of 316,959 new Covid tests.

Florida’s testing got off to a very slow start until March 13th, when 195 tests were made in one day. In total, Florida has a daily average of 5764 tests a day.

This average is a little misleading because it took a few weeks for Florida to get up to speed. So I calculated only the last 23 days, when the state reached its first 10,000 tests in a single day.

Using this time span, the state average 9724 dally tests. This is about the current daily average, but is still significantly below what is needed to meet the minimum levels some that epidemiologists find acceptable: 32,000 daily tests.

But how is the state doing according to the 10% rule? Using this metric, Florida is doing alright. The lastest testing data shows that Florida is right at the 10% level, as shown in the graph below.

The blue line represents the percent of all tests since March 5th. As you can see, the state’s positive rate started out high, then dropped quickly in the first few days. But the testing rate during this point was quite low until March 13th. As the number of new tests ramped up, the percent of positive results skyrocketed to over 20%.

By the end of March, however, the positives began to stabilize as the number of tests also increased as shown by the following graph.

No you are not seeing double! The number of positive tests by the total number of tests is identical to the graph by date. The number of tests increased as time increased.

But the message is the same, positive Covid-19 tests fluctuated as the testing increased and then stabilized around the 10% level. In other words, when increased testing reached a certain level, in this case around 145,000 total tests, the virus’s natural Florida infection level is 10%.

The importance of having a low positive testing rate is in it’s relationship to future deaths. As we all know, no new Covid-19 cases leads to no new deaths. A graphic representation of what happens when the number of positive Covid-19 tests increases is shown below.

Using the Covid-19 Tracking Project testing data, our Cubic Regression Model displays the almost exponential increase in deaths as the number of positive tests multiply. The model predicts that 3.4% of new positive Covid cases will eventually succumb to the virus (US total is 5.7%). That’s three percent too many.

So how is Florida doing when it comes to testing? The national positive test rate is about 20%. So compared as the nation, Florida is doing well in comparison. But it’s not there yet.

If we are going to get back to working, drinking and eating out again we need to significantly increase the number of tests per day. That number is probably closer to at least 20,000 tests a day.

While we are waiting for the vaccine, we need to know who has the virus, who has had it and who has it but has no symptoms. Otherwise, we are fighting this battle blind.


Good News Governor DeSantis says we have “Flattened the Curve”

It’s obvious the Governor doesn’t follow this blog, otherwise he would have discovered that Florida’s curve had already flattened on April 13th. It’s probably just as well, since he may have wanted to open up Florida back on April 13th. This state will not be safe to open until we have eliminated the Covid-19 new cases.

Our Covid-19 model continues to show a slow downward progress of new cases, as shown below:

Data as of April 21 , 2020

As you can see by the red track line, our model continues to predict that the number of new cases will continue to decline. Florida’s new cases declined for three days, but yesterday it increased by 150. This daily variation is normal and not a trend according to the model.

As expected, the number of new daily deaths continues to rise, as shown in the graph below:

The trend line continues its exponential movement skyward. As you will see, the national model shows a different pattern. The difference is more likely due to the number of national deaths as compared to that Florida model. The national data has far more daily deaths than Florida. A larger sample tends to spread out the data. The national model of new cases is not as far along as Florida as well, displayed in the next graph.

The US model shows the nation is not as far along in the curve, as the Florida model. But the model clearly shows that the nation is on its way downward. It looks like the US is about a month behind Florida in its control of the virus. But the number of daily deaths cases continue to rise, as graphically shown below.

Again, because of the lag time (this is the period between infection and death), new deaths will only decline until the number of new cases significantly falls.

The good news is that new Covid-19 cases continue to slowly fall in both Florida and the United States. The bad news is the number of new deaths continue to rise!

Unfortunately, some governors are expected to cancel their “stay at home orders” and open some more essential businesses such as tattoo and massage parlors! That will certainly set the economy on fire…skip the massage and stay home.


The Incredible Shrinking Biden Lead?

You probably noticed that we don’t see that much of Joe Biden since the U.S. virus epidemic. And you have almost certainly seen Donald Trump every afternoon for his so-called Covid-19 virus task force briefing.

I have already shown that the recent Covid virus epidemic in the Nation has had no effect on Trump’s Job Approval numbers, so it didn’t dawn on me that it could effect the Biden/Trump two party vote distribution. In other words, the head to head polling between the two candidates.

But a cursory look at the last couple of months of polling, it looked to me that Biden’s early and consistent lead in the trial ballot question was narrowing as time went on.

So I decided to take a closer look at all polls that included the Biden/Trump, “if the election were held today” question. I found 102 surveys from March 17, 2019 through March 18, 2020 that met that criteria.

The average of all 120 polls showed Joe Biden with 50.5% and Donald Trump at 42.3% (balance undecided). Of all polls, that is an average lead of 8.2% for Biden.

But as you will see in Chart 1 below, this average has slowly eroded in the past year and a half.

Chart 1

The red line on the chart is the percentage of Trump’s percent support from each survey during the past year and a half. The green line represents the percentage for Biden.

The lines zig-zag up and down reflecting the variability of each poll result. But if you look closely, you will see the direction of each line, the red staggers upward and the green in the opposite direction. Over time, the green line (Biden) shifts downward, while the red line shifts upward.

A clearer picture of how Biden’s significant lead has evaporated over the last year and half, is the net percent lead that Biden has had over Donald Trump, as graphically shown in the Chart 2 below.

Chart 2

The blue line represents the polling difference between Trump and Biden’s percent of the polling vote (Trump – Biden %).

What is obvious from the chart is how the blue line gradually zig-zags up to the zero horizontal line. In other words, Biden’s percent lead over Trump has shrunk significantly over the past year and half. (ANOVA sig. <.06 level).

In the last six months of last year, Biden’s average lead over Trump was net 13.5%. Since recent Covid-19 epidemic it was 6.1% and continuing to shrink. In the last six polls, Biden’s average lead was only 2.4%.

Now there are probably several reasons for this percentage decline. I think we can rule out campaign effects, since both campaigns have gone pretty much underground, with even Google refusing to take political ads now.

Since most of the decline has occurred since the first cases of the Covid-19 infections, I would suspect some of this decline is due to the recent epidemic. And I don’t mean that voters don’t like Biden’s handling of the crisis.

Instead, I posit that its not the virus, but Biden’s visibility has been overwhelmed by the news coverage of the epidemic. As my mother used to say, “out of sight, out of mind.”

This is not a planned break, but the inevitable outcome of being overtaken by a crisis that has consumed every TV set in America. Trump, on the other hand, is front and center, mainly praising his handling of the crisis.

Don’t misunderstand this point: Trump’s job approval rating hasn’t varied 2% since the Covid-19 virus took over the U.S.. But he has the stage and Biden has an occasional side show.

Finally, the recent polling data underscores that this presidential race is likely going to be close. Hopefully, the virus starts to wane and politics will again occupy the American voter’s attention. Elections do matter and this election matters even more.


Todays Daily Covid-19 Models

April 20, 2020

Our latest Florida Covid-19 model continues its predicted downward trend for new cases. Data near the top looks a little messy, but that suggests that near the peak the daily number of new cases goes up and down as it adjusts to model’s expected path.

For the last three days the number of new Florida new cases have declined. The model’s predicted path is still strong and shows no signs of deviation. Good sign.

The number of covid-19 deaths, unfortunately, continues to climb unabated, as shown in the model’s graph below.

The number of new deaths is almost a straight line upward. With no significant curve in sight. As you will see below, the national picture shows some of rays of hope that might herald a decline in the near future.

The model clearly shows that the Nation is at the end of the curve and is starting it’s trek downward. The number of new cases have declined for the past three days. Unlike Florida, the nation’s new deaths may be beginning it’s curve as well, according to the model’s estimate.

For the last two days, the number of new Covid deaths have declined by 15%. Just a reminder, however, although the path of the virus in the model is clear, the daily data points will sometimes bounce around. That is expected, but the predicted trend path has not shifted.

Check in on a regular basis for model updates. For now, we just have to wait out the virus. Opening up the state before we are near the bottom would be a mistake.

With a estimated spread rate (Ro) of 2.3, it wouldn’t take long for the state to spike upward again. I don’t know about you, but I want this isolation to end, but not at the expense of losing all that we have gained during the past two months. Stay home and be safe.


Florida Covid-19 New Cases Slowly Moves Downward

April 17, 2020

The virus is trying to stay alive, but the direction is still downward, as pictured in the model graph below.

Yesterday’s number of new cases spiked to 1200 from the previous day’s 941 cases, a 27% increase in one day, but the overall model’s trajectory continues to point a slow downward movement. These kind of spikes are not unusual so don’t panic yet.

Based on the President’s new virus guidelines for opening up their economies, suggests (requires?) a 14-day downward trajectory in documented new virus cases. It does not specify any number or percentage of decline, just that the trajectory is downward.

Under that criteria, Florida should easily meet that requirement by the end of April. Taking any action is left up to the governor of each state. Personally, I think there should be a maximum number of new cases or at least a maximum percent of new cases to begin the process. In the end, however, it’s up to the governors’ discretion. The President doesn’t want the ball…

As expected, the number of Covid-19 daily deaths have continued to rise, as shown below.

As long as we have new cases, there will be some deaths. But as the number of new cases decline, eventually the number of new deaths will decrease as well.

The national level of new Covid-19 cases continues to decline, but a slower pace, as shown below.

Some of you have seen on TV recent graphs have shown the number of cases rising the last few days.

These are simple bar graphs used to measure linear data. Our model is a cubic regression model, which is more appropriate for the virus. As the new cases are added, the model adjusts the trend line accordingly. One or two days is not a trend.

The model’s direction continues to predict a continued downward motion, but the pace is slower than expected. The last few cases have clustered in a small group but still in a downward motion. Remember that this is a national model and many states have different mitigation policies and the effects of these different policies will have small but diverse outcomes.

The nation’s daily deaths, as you would expect, continue to grow, as displayed in the new graph.

Notice the larger red circle in the upper half of the graph. That represents 6185 deaths in a single day! This unusual spike occurred on April 14. The next day it dropped to 2700 deaths and slowly declined from there.

Why that spike? I really don’t know. It’s probably deaths that occurred earlier but for some reason, were dumped on this date instead.

A large number of nursing home deaths have been occurring nationally. It’s possible that some of these deaths are from nursing homes that have been holding back on notifying their state health departments. If I find out, I’ll pass it along. In any case, this anomaly only marginally effects the model and not the continued direction.

It’s apparent the social distancing requirements are having significant effect on the spread of the virus. That said, the states’ will be walking a tightrope, balancing the virus containment and the desire to get back to normal. Good luck with that…

In the end, some people will be unhappy that we are either not opening up fast enough or opening up too soon. President Trump’s decision to let the Governors’ make the decision how to handle the problem was politically smart.

The new issue is now testing and most experts claim that opening up without adequate testing will likely lead to a resurgence of the virus. Trump has made it clear that it’s the states’ problem.

He hand’s the ball to the Governors’ and then tweets to his base: “liberate” Michigan and Minnesota, putting pressure on those governors’ to open up.

If new cases and deaths spike, it wasn’t his decision, those Governors’ wanted to handle this themselves. If the states go too slow, he could of have done a better job. How can he go wrong.

Be safe and stay at home if you can…


Florida Continues its Curve Downward

April 15, 2010

The number of new Florida Covid-19 cases continue to decline as our model below shows.

Data as of April 14

Yesterday’s new cases made a significant drop from 1002 to 614, or a 63% decline in one day. Unfortunately, the daily number new deaths are continuing to more upward, from 524 to 591 deaths. Deaths are likely to increase until the number new cases significantly declines, remembering the lag time for Covid deaths is approximately 13 to 18 days from infection.

The national picture is also improving as well, albeit at a slower pace Florida, as shown in the model graph below.

Like Florida, the number of U. S. new cases are also falling as well, along the models projected path (red line). The confidence level of the model is high, with an Rsq. of .973.

Like Florida, the number of national new deaths are still climbing with no end in sight. Eventually, we should see a leveling off on new deaths, but that will depend on new Covid cases declining significantly.

So when can we see an end to new cases? For Florida, the model’s track suggests that the number of new cases will near zero (but not zero) between 45 and 55 days from the beginning cases, which indicates an April 20 through April 30 date. Let’s say May 1st. For the nation, the date is between May 1st to May 15th, if we continue social distancing and stay at home orders.

Deaths, however, will continue to increase for a few weeks because of the lag time between new cases and death. Even after the last of new cases reach zero, deaths will continue to occur.

I’m not as confident in the end dates, since the key variable is human behavior, which is always unpredictable. When we reach those dates, I’ll evaluate the time estimate. Be safe…and stay at home if you can.


Has the Coronavirus Hurt Donald Trump’s Reelection Chances?

April 14, 2020

With the economy tanking and the coronavirus’ deaths rising, some insiders are beginning to wonder if President Trump’s reelection chances are declining. The Tampa Bay Times recently polled a 160 political insiders and 80% said the Coronavirus has hurt Trump’s chances for another four years in the White House.

To be candid, my own impression of the crisis coincides with these insiders. Trump has made the economy his silver bullet. But as I look from out of my home office window, isolated from the virus world, I wonder if there is an economy anymore and if it will come back by the November.

In addition, his handing of the virus crisis is under attack every day by both the media and the Democrats. Although the President hold’s a TV show every afternoon, even Republicans are wondering if these “press conferences” are hurting or helping his election chances.

But is the Covid-19 virus really hurting him at this point in the election cycle? A recent post in FiveThiryEight (April 14) showed that 47.7% approved of his handling of the crisis, while 48.5% disapproved.

Many of your know my affection for the President’s Job Approval rating as a metric of the President’s popularity. It’s importance is highlighted by the fact that no incumbent president since World War 2 has ever won reelection with an approval rating below 50%. That applies, of course, only to approval ratings just before the election.

Some pollsters have noted a slight rise in Trump’s rating near the end of March, but it quickly returned to its’ mean (around 45%) soon after. Some have attributed this bump to what political scientists call the “rally around the flag” effect, which often occurs around periods of crisis.

So to see if the virus crisis and the economy’s collapse has hurt or helped Trump’s approval rating, I collected 47 days of approval ratings as posted on Real Clear Politics and FiveThirtyEight for every day since January 3. At least one approval poll occurred on each day, with most days having at least 3 to 4 surveys with this metric. When more than one poll occurred on a day, I averaged the rating of all polls for that day.

To determine if there were differences from days before the first virus case and those approval ratings after the US cases started occurring, I split the data into two samples: pre-coronavirus and post-virus periods. The pre-virus cases will act as our control group. (January 3-March 3)

Our control group (pre-virus period) has a job approval mean of 44.2%. The mean for the post-virus group is 45.7%, or a 1.5% difference. To determine if this difference is significant, I used Analysis of the Variance (ANOVA) which is used to determine if there are differences between two groups (pre-virus vs. post-virus). A graphical representation of how similar these two series look like are shown in Chart 1 below.

Chart 1: pre-virus pols (red line) and post -virus polls (blue line)

The ANOVA results shows that the statistical difference is not significant (.675) between the two groups. In other words, there is no difference between the polls before the virus epidemic and those conducted after the Covid-19 cases and deaths became public in the US.

This result also negates the idea that there was a significant bump in Trump’s Job Approval when the initial covid-19 cases occurred that many have suggested was a “rally around the flag” effect. It was just a random bleep.

All of this surprises me, but has confirmed my inability to find trends in Trump’s job rating as the number of new cases rose. Initially, in times of crisis, voters generally support their elected leaders. That has not happened here. As the crisis worsens, voters often recant their initial approval. That hasn’t happened yet either, and I’m not sure it ever will.

My only explanation is that the Trump base is made of granite. And the non-Trump group’s negative ratings are not abating either. Each side has made up their mind and even an endorsement from the Lord Himself isn’t going to change people’s opinions. God Bless…


We are headed down the Curve!

April 13, 2020

Mark this date down on your calendar, it’s the day we started toward normality. Our latest model shows that Florida has peaked and begun the long trek downward to no new virus cases, as shown in Chart 1 below.

Chart 1

Notice at just past the peak of the curve (red line) there is a cluster of cases on the early downside. This is an early sign of q decline in new cases. There are of couple of outliers, but they are new cases that are even lower than the cluster cases. All point toward the beginning of fewer cases.

That doesn’t we mean are out of the woods yet. There still could be few bumps along the way, but it all points to a downward trend. That is, if we continue social distancing.

There is a similar pattern developing at the national level as well, although slightly behind Florida, as shown in Chart 2 below.

Chart 2

There is a clear peak here as well and the beginning of lower data points (new cases). Like the Florida model, the red trend line predicts a steady decline in new cases.

Although this very good news, it’s the just the beginning. of long process to normalcy. Stay at home and wear a mask when out, or you still end up on the chart below: Covid-19 daily deaths.

Unfortunately, the number of Florida daily Covid-19 deaths continue to climb. As I have mentioned in previous posts, the time lag between new deaths and new cases is approximately 13 to 18 days from infection. Deaths will likely continue to rise until there is a significant decline in new cases over a few weeks.

The same is true for the national virus epidemic as well, as demonstrated in the model’s estimate below.

National daily deaths, continue to climb as well, but unlike Florida, there are some early signs of slowing down, which is surprising. Yesterday, some 1528 deaths in the US. The day before, there were 1830 deaths recorded. This could be just a random event, but the next few days will confirm or reject this early trend. I will keep watching it….

What is the Doctor Politics Coronavirus Model

The basis for using a polynomial model is that the data you are exploring is non-linear. Meaning that one data point is not a linear relationship to another data point. In the case of new virus cases, we would not expect the number of new cases to move solely in a straight line. Our expectation (hypothesis) is that as the number of new cases increases, it will likely hit a plateau and eventually slow down and then decline, making curve at the top as it progresses.

This assumption requires a curvilinear model to track that change. This requires a method that follows the “curves.” Normally a quadratic regression model would be used, but that model only allows for data that curves in one direction and in a U or convex pattern. In the case of a virus, the data could wander in both directions (convex and concave).

The best method for this type of data is called a cubic regression model, which allows for data that can curve in both directions (convex/concave) and it uses three cubed predictors.

In the simplest of terms, the model can track the non-linear virus data points (a curve), and use that movement to predict the virus’s likely future path. Cubic regression has been used to explore complex DNA changes and in sophisticated stock market models.


Florida New Cases Beginning the Curve?

April 11, 2020

After stalling at the top, Florida’s new Covid-19 cases looks like it’s starting a slight downward slope. The Chart below shows our polynomial trend line still bending downward as new data fills in behind it.

We should see more confirmation in the next few days. But for now, our model continues to point toward a coming decline in new Covid cases. Deaths, however, continue to increase due to the lag time between new cases and and death.

The CDC estimates, based on data from China, that from the onset of pneumonia to death is 13 days. Symptoms of Covid-19 usually start within five days from infection. Using that range, we can broadly estimate that from infection to death would take approximately 18 days from contracting Covid-19 and death. Death, of course, is still relatively a rare event, since the estimated Case Fatality Rate (CFR) is only 3.4%. For every 1000 new cases, results in 34 covid-19 deaths.

The US numbers also appear to be near a peak in new cases as well, as shown in our national model below.

The trend line is also bending downward but the data points are still just below the expected peak. We need more confirmation that the number of new cases has declined before we can say the peak has arrived nationally. But the signs are encouraging. As you would expect, the number of new deaths continues to significantly out pace the number of new cases.

I will continue to post these estimates daily, so keep checking.


Florida Has likely Peaked

Friday April 11, 2020

Today’s Florida coronavirus model strongly suggests the state has reached or nearly reached the peak of the virus epidemic, as shown in Chart 1 below.

Chart 1

Over the last few days, new Florida Covid cases have clustered at the top of the trend line (red) of our exclusive polynomial model (cubic). The model is expecting a trend downward, suggesting that Florida’s new cases will soon begin to decline.

The national numbers show that the peak hasn’t arrived yet. But the model’s projection, continues to suggest that a peak is coming.

In both Florida and the nation, the number of deaths are still climbing reflecting the lag time between new cases and death. As the number of new cases decline, we should begin to see a drop in these numbers as well. Be safe and keep watching for future estimates.