SADOW: Those Virus Testing And Death-Rate Numbers Don’t Add Up

https://thehayride.com/2020/04/sadow...lJWxeG8vT2SpOc

SADOW: Those Virus Testing And Death-Rate Numbers Don’t Add Up

With little attention or fanfare, this week Democrat Gov. John Bel Edwards extended a series of proclamations limiting gatherings and commercial activity to stave off the Wuhan coronavirus pandemic. He did this using data that appears increasingly questionable.

Observers such as MacAoidh already have pointed out not only errors in data collection that affect perceptions about the disease’s contagion in Louisiana, but also have noted timing involved in releasing specific numbers on specific days fits a pattern designed to produce messaging that promotes a political agenda. Certainly, continuation of the Edwards bans throughout the entire state idles further a Legislature chomping at the bit to enact policy contrary to Edwards’ liking while it embolden actions comporting to his general philosophy of expanding government through increased taxation and spending.

The extension again raises the question of whether a blanket approach serves the state’s best interest. The figures for today continue to show the virus, at least in a critical form, remains largely a New Orleans-area phenomenon. Orleans, Jefferson, and their surrounding St. Tammany, St. Bernard, Plaquemines, LaFourche, and St. Charles Parishes have but 28 percent of the state’s population but 69 percent of the cases and 73 percent of the deaths.

If you can trust these numbers. Whether through shoddiness or deliberate manipulation, it has become obvious some reported numbers strain credulity, if they make any sense at all. It begins with deaths.

The introduction of the virus came in Washington, first in Snohomish then King Counties. With its initial appearance and particularly in a convalescence hospital, King has a high death rate from infection, at 6.64 percent as of yesterday. With King having a third of the state’s population, this has driven the statewide rate to 4.13 percent. That’s higher than the rate in Orleans as of today at 3.97 percent, and statewide of 3.39 percent.

The area surrounding New York City (itself comprised of five counties often known by their borough names) – Nassau and Suffolk Counties on Long Island, Westchester and Rockland Counties to the north on each side of the Hudson River, and the New Jersey counties across the Staten Island Sound of Bergen, Passaic, Hudson, Essex, Union, Middlesex, Monmouth, and Ocean City, with the city since have become the cockpit of the virus. But within this area of about 18 million people, which contains just under half of all U.S. cases and four-ninths of its deaths, it has a death rate only half that of Orleans.

And deaths per capita tell an even more extreme story. Consider that King and its suburb Snohomish have registered 203 deaths among their more than 3 million residents, just five more than Orleans and Jefferson that have just over 800,000 people. King has 75 deaths per million from the disease, while the New York City area has registered 111 per million.

Yet, stunningly, Jefferson rings up a death per capita figure more than twice King’s, at 168 per million, and Orleans a number almost three times that of the New York City area at 320 per million. This likely is the highest such statistic among large metropolitan counties in the U.S. – but not the highest in Louisiana, where upriver St. John the Baptist Parish data compute to an amazing 322 per million.

Throw in the parishes surrounding Orleans and Jefferson and the parish upriver from St. John the Baptist (which itself is just upriver from St. Charles), Ascension, with those six together having a rate of 58 per million, and it becomes clear why Louisiana has a statewide rate of 67 per million. Not only is this double Washington’s, it’s higher than all of New Jersey and trails only New York’s 121, supercharged by New York City’s number of 161.

Only two things can explain why Louisiana has a death rate per capita from this malady far exceeding the national average of 13.8 per million: something about its people and/or health care system makes Louisianans much more likely to die from this disease, and/or the state reports deaths in a way that inflates the cause being the virus.

Then there’s the testing numbers, which, in two words, are entirely unreliable. According to state data, Louisiana in aggregate has had just over 51,000 tests performed, representing 1.1 percent of the population. This is midway between the Washington figure of 1.02 percent and New York’s 1.22 percent, although well above New Jersey’s 0.66 percent. Positive tests registered in Louisiana about 18 percent of the time, as opposed to 8 percent in Washington (which had an aggressive program from the start), New York’s 38 percent, and New Jersey’s 43 percent (both of which, like Louisiana, started relatively late).

But the proportions vary widely among parishes, and in many cases this seems likely a consequence of counting error. For one thing, as of today the individual parish totals didn’t add up to the state numbers. State statistics showed 3,901 conducted by the state and 47,185 by commercial entities. Yet the sum of parish totals had 280 fewer from the state and 447 fewer done by private labs.

Worse, some parish numbers seemed unlikely. St. John the Baptist, with its infection rate of 1:159 not much higher than Orleans’ 1:124 (which is the highest in the nation outside of the sub-100 numbers from Rockland and Westchester Counties in New York), had its 274 cases identified from just 25 tests – a mathematical impossibility unless commercial testers reported from their location and not the parish of residence of the person tested, if that person sought a test from outside his home parish. Four other parishes had more cases than tests, or close to that.

In other instances, the sheer volume and tests and proportions positive raise eyebrows. Jefferson has rung up 2,178 positives from just 3,077 tests. Again, maybe along with St. John residents they are running to Orleans to have their testing done (over 23 percent of tests done there have turned out positive), but that makes for some weird dynamics: why would testing facilities in Jefferson disproportionately attract people with it and/or have those residents without it have their testing done elsewhere?

And if that’s how it works, categorizing in this fashion doesn’t do a whole lot to produce data useful for decision-making about the pandemic. For example, in a radio interview Edwards implied a rapidly climbing number of cases popping up in Caddo and Bossier Parishes meant these citizens didn’t behave carefully enough, but that doesn’t square with the two having more testing done by far than any single parish except for Orleans and a relatively miniscule positive rate a quarter of Orleans.’

In Thursday’s news release about the most recent numbers (which apparently skewed upwards because previously unreported cases from days ago finally made it into the official tally,” Edwards noted that “I have said time and again – COVID-19 is a statewide problem and testing is a vital step towards understanding the scale of this problem.” In reality, it mostly is a New Orleans area problem, and the way testing data is presented confuses rather than clarifies the scale of the problem.

LOL, Dilbert Why do conservatives arguing about this dangers of this virus?

LOL, Dilbert Why do conservatives arguing about this dangers of this virus? Originally Posted by FrankZappa

It is dangerous to old and unhealthy people.

New Orleans has a bunch of fat and unhealthy people who have died a few months earlier than something else that would have gotten their lazy and unproductive asses.

They should be quarantined and the rest of us put on masks and get to work.

Delbert,

I appreciate your concern. The numbers don't add up because they are not suppose to add up.

What? Because this is not simple addition and subtraction.

Modeling examples (read that as the dirty word statistics) are projection of worst case predictions of what can be, or will be, without action.

We are not talking about describing what things are with numbers, but prediction what thing could be in the future. With those numbers we will attempt to change the future. Been there and done that. In the beginning college courses we talk about the standard curve, usually used to show something about the population, or a trait of the population - such as intelligence. As used, those curves are manipulated to achieve a far different outcome. ie, the curve in now changed to be what we want, it is not an average - it is not normal.

As applied, statistics and models attempt to change that standard curve or to change what will occur.

The numbers told us about how many dead we could expect with a freight train with a dangerous cargo running down the tracks without anyone in the engine and heading for a city full of people. If it reaches the city, many will die.

What would you do with that knowledge? Easy, stop the train by any method available. Derail it: What? Don't have a dearailer - tear up the tracks. Don't have time block the train with another, or put a bunch of trucks on the rails. Got any explosive? Blow up the tracks. Don't have explosives, think of something else. What? First, get to thinking. Is there an Air Force Base near enough to scramble the plains and bomb the tracks?

Don't like the RR example, take this. Andrew Carnegie, the founder of US Steel, was working for the RR as a young man when the main managers could not be reached. There was a train wreck with an express train due to come through the wreck area. On his own authority, he sent a telegram to burn the wrecked train and get it off the tracks. After that he was rapidly promoted.

The point, in a time of crisis we can not wait to make decisions. Decisions have to be made immediately. President Trump make the decisions and got it done. Quit your bitching: he changed the outcome. Most other politicians would have worried about how they would be viewed. Trump was the boss, and he followed the advice of the experts.

Got the idea, We know what will happen if we do nothing, now that we have slowed the train down, quit complaining that the predictions of it causing a lot more deaths isn't as stated in the beginning.

It is predictive, and with that we change the outcome.

This is like Walmart using statistics to determine how many check out lanes it will need for each day of the year. Hey, do we have records on how much is sold each day of the year? Think about the Christmas season.

Yes, we have the statistics and we use them to put more clerks in the check out lanes. Now you know why most of the year stores have extra checkout lanes they are not using. Can you also know why we have plans inn place to handle problems like the virus. The problems is that some politicians (Read that as mostly Democrats, but not all some Republicans are stupid too.) do not see the need to be prepared when there is not an immediate payoff.

Another example: Every day there is crime in a given spot of the city - over and over again. We know that, we keep records. Then we put more police at this spot, every day. Now the crime goes down, so don't cry about the cost of the added police at the spot targeted. Oh shit you say, it just moved somewhere else. Guess what we have records for that somewhere else too, and everyplace else.

That is how models are made, and changed.

A long time ago, I collected the records made the models and statistics for a company, with over a dozen food production plants, for both energy consumption, production/packing and quality control statistics.

We were overfilling each package. Each package lost weight over shelf storage, but was still overweight after one year. Question: How much of the overfill weight could we reduce? Answer $2.5 million/year. That is every year from then on, or $25 million in 10 years.

We were using electrical compressors with surcharges for excessive usage during startup. Question: How to reduce the electrical cost of refrigeration. Answer: Keep a steady refrigeration cost by adding more insulation, keeping the refrigeration steady in cold storage, and staggering the production refrigeration startup to maintain a steady use of the electrical compressors. ie, bring in an extra mechanic early to stager the compressor startup and pay him extra for the overtime in each plant. Net savings: slightly over $1.3 million/year.

Project time: I took me a 10 months to run the two studies and 2 months to get the plans put into place at all of the plants. My reward for a job well done: $75/month, it was my job: what did I expect. They thought I was just doing a job. What? I looked for another job and put the project on my resume to get much better pay.