A writer in Science says we must form an integrated multinational program led by some select group with massive funding to have a Manhattan Project for COVID-19. Specifically he notes:
An initiative of this scale won't be easy. Extraordinary sharing of
information and resources will be critical, including data on the virus,
the various vaccine candidates, vaccine adjuvants, cell lines, and
manufacturing advances. Allowing different efforts to follow their own
leads during the early stages will take advantage of healthy competition
that is vital to the scientific endeavor. We must then decide which
vaccine candidates warrant further exploration purely on the basis of
scientific merit. This will require drawing on work already supported by
many government agencies, independent organizations like the Coalition
for Epidemic Preparedness Innovations, and pharmaceutical and biotech
companies to ensure that no potentially important candidate vaccines are
missed. Only then can we start to narrow in on those candidates to be
advanced through all clinical trial phases. This shortlist also needs to
be based on which candidates can be developed, approved, and
manufactured most efficiently.
Unfortunately the author in my opinion is clueless. The Manhattan Project was a secretive single focused technical program to implement a weapon. The COVID-19 effort is a highly distributed effort interconnected in an open manner examining a multiplicity of approaches and leveraging off the open market capabilities of many entities. Any singularly focused and led effort of an international scale will be a disaster. Just look at the WHO. Their gross incompetence, in my opinion, frankly led to this.
Tuesday, March 31, 2020
Models and Complexity
The NY Times alleges that the Government is to release the models today. We frankly have no idea what that means. The description, the results, the source code, the underlying methodology? Since it is the Government they also have no idea what they are doing.
Let's give a simple yet complex example. Namely how does one model the transfer of the virus. Consider the example below:
Here we have an example of person to person infection. The infected individual can infect an uninfected by two means. First by aerosolizing the virus and transferring by air directly. That could be stopped by a mask. However proximity and time continuity is essential for this path to function. Second, by secondary touching. Namely the virus transferred say by hand to a surface, then the second person must touch the surface at the right spot in the right time period and transfer to their hand and then to their face to start the process. Gloves and hand washing stops that transfer.
We now above show a model for this process. Namely each step has a probability and each step has a mitigation element. We can then estimate the probability of infection from one person to another at a specific time period in a specific table surface at a specific temperature. Lots of assumptions.
Then if we want a model we have to assume that there exists an ensemble of people al functioning in some statistically similar manner. I think one starts to get the point. Assume, assume, assume etc
The issues continue since how this happens in a NYC subway is not how this happens in Scranton, and less so say in Bangor. But, and this is critical, this simple example is at the very heart of the model.
Problems like this were considered a century ago when examining statistical thermodynamics. It became an element of Einstein's Nobel Prize winning paper on Brownian Motion. Unfortunately the application to pandemics is grossly wanting.
As the above NY Times article states:
The models used by the White House team are standard epidemiological tools but are not precise, as the results can vary widely depending on how closely people follow the guidelines. In other words, the assumptions built into the models can shape the results.
This is an understatement. Having designed and tested such models and examined hundreds of others, and also understanding some of the dynamics of this virus, the simple answer is that no one has a real clue. Yet the two steps shown above of distance and hand protection are the best we can come up with now. My grandmother, Hattie Kruger, told me this seventy plus years ago from here experience in dealing with TB and the Spanish flu in New York. She lived to 90, so believe your grandmother. Thanks Hattie!
Let's give a simple yet complex example. Namely how does one model the transfer of the virus. Consider the example below:
Here we have an example of person to person infection. The infected individual can infect an uninfected by two means. First by aerosolizing the virus and transferring by air directly. That could be stopped by a mask. However proximity and time continuity is essential for this path to function. Second, by secondary touching. Namely the virus transferred say by hand to a surface, then the second person must touch the surface at the right spot in the right time period and transfer to their hand and then to their face to start the process. Gloves and hand washing stops that transfer.
We now above show a model for this process. Namely each step has a probability and each step has a mitigation element. We can then estimate the probability of infection from one person to another at a specific time period in a specific table surface at a specific temperature. Lots of assumptions.
Then if we want a model we have to assume that there exists an ensemble of people al functioning in some statistically similar manner. I think one starts to get the point. Assume, assume, assume etc
The issues continue since how this happens in a NYC subway is not how this happens in Scranton, and less so say in Bangor. But, and this is critical, this simple example is at the very heart of the model.
Problems like this were considered a century ago when examining statistical thermodynamics. It became an element of Einstein's Nobel Prize winning paper on Brownian Motion. Unfortunately the application to pandemics is grossly wanting.
As the above NY Times article states:
The models used by the White House team are standard epidemiological tools but are not precise, as the results can vary widely depending on how closely people follow the guidelines. In other words, the assumptions built into the models can shape the results.
This is an understatement. Having designed and tested such models and examined hundreds of others, and also understanding some of the dynamics of this virus, the simple answer is that no one has a real clue. Yet the two steps shown above of distance and hand protection are the best we can come up with now. My grandmother, Hattie Kruger, told me this seventy plus years ago from here experience in dealing with TB and the Spanish flu in New York. She lived to 90, so believe your grandmother. Thanks Hattie!
Labels:
Pandemic
Monday, March 30, 2020
Three Card Monty
There are a set of "models" out there which are driving policy. I suspect that every politician is clueless and thus just take the words as if sent down from Delphi. On the other hand it is akin to a three card Monty game on the streets of NYC. Shuffle the cards, place your bet, guess where the ace went!
These "models" should, nay must. be made public, so that we can all see the assumptions. For example in the simple model above we assume we do not know a(t), such as a parameter in the pandemic model. We then apply a system identification routine to estimate a(t) based on the actual data as it is gathered and then adjust the model accordingly. I did this for Apollo guidance over fifty years ago, and my first book covered this. I suspect these models do not adjust accordingly.
The current scheme is we have some wizards secreting a model and whispering the result. NIH and CDC apparently do not have this expertise, nor would one expect in NIH, but CDC?
The solution is simple, set the models loose, let any and all who want to take a shot at them do so. Also we need the raw data in a readily useful form. The problem is that anyone truly competent in the field of dynamic estimation of systems with unknown variables can and should be involved. Otherwise it is a three card Monty game, no one trusts the numbers. Worse, we have experts in other fields mouthing off as if these numbers are from the mount. They are not.
These "models" should, nay must. be made public, so that we can all see the assumptions. For example in the simple model above we assume we do not know a(t), such as a parameter in the pandemic model. We then apply a system identification routine to estimate a(t) based on the actual data as it is gathered and then adjust the model accordingly. I did this for Apollo guidance over fifty years ago, and my first book covered this. I suspect these models do not adjust accordingly.
The current scheme is we have some wizards secreting a model and whispering the result. NIH and CDC apparently do not have this expertise, nor would one expect in NIH, but CDC?
The solution is simple, set the models loose, let any and all who want to take a shot at them do so. Also we need the raw data in a readily useful form. The problem is that anyone truly competent in the field of dynamic estimation of systems with unknown variables can and should be involved. Otherwise it is a three card Monty game, no one trusts the numbers. Worse, we have experts in other fields mouthing off as if these numbers are from the mount. They are not.
Labels:
Pandemic
NJ 2020 03 30
The infected continue to climb, the rate of increase in still 20-30% per day.
The day by day changes per county are below:
Overall there is an increasing rate. Part of the problem may be delayed or improperly timed reports.
The total per county are above. Bergen dominates. But from Somerset down there is almost negligble counts and growth.
On a per PoP basis by county above we see the bottom 4 are truly almost lacking in infection.
The above are the percent increase normalized. They are somewhat consistent and thus independent of PoP density. This is a classic modelling issue.
The above are the daily normalized changes by county. The Bergen numbers are startling, PoP density is lower than Union and Hudson.
The day by day changes per county are below:
Overall there is an increasing rate. Part of the problem may be delayed or improperly timed reports.
The total per county are above. Bergen dominates. But from Somerset down there is almost negligble counts and growth.
On a per PoP basis by county above we see the bottom 4 are truly almost lacking in infection.
The above are the percent increase normalized. They are somewhat consistent and thus independent of PoP density. This is a classic modelling issue.
The above are the daily normalized changes by county. The Bergen numbers are startling, PoP density is lower than Union and Hudson.
Labels:
Pandemic
Sunday, March 29, 2020
Some Thoughts on the Pandemic
In the field of marketing one looks to get a target market. To do so we need both demographic and psychographic measures as who in turn has a propensity to buy so that we can then promote and persuade this group. By understanding their demographics and psychographics we know what to say and where to say it.
Now with COVID-19 we can apply the same techniques in a rather simple manner. If we know the infected we then just use their social media train. From that we get both elements plus we get all their contacts. We can then cross tab the results and find out the clusters and better focus our approach. Perhaps it is a male, 23-35 years of age living in Bushwick who takes the A train to 42nd street each day. Just a guess. But the data is there. Instead of fallacious epidemiological data methods, often rant with useless assumptions and erroneous parameters, we get targeting data based upon current factual relationships and understandings.
Then we can reverse engineer the social media platform to target the next anticipated infected to hopefully quasi-quarantine them. Thus instead of a sledge hammer a la Trump, we can use our brain and data readily available to focus on remediation.
Just a thought for the day. Probably will not work since these platforms are not know to be supportive of the prosperity of the US population.
Now with COVID-19 we can apply the same techniques in a rather simple manner. If we know the infected we then just use their social media train. From that we get both elements plus we get all their contacts. We can then cross tab the results and find out the clusters and better focus our approach. Perhaps it is a male, 23-35 years of age living in Bushwick who takes the A train to 42nd street each day. Just a guess. But the data is there. Instead of fallacious epidemiological data methods, often rant with useless assumptions and erroneous parameters, we get targeting data based upon current factual relationships and understandings.
Then we can reverse engineer the social media platform to target the next anticipated infected to hopefully quasi-quarantine them. Thus instead of a sledge hammer a la Trump, we can use our brain and data readily available to focus on remediation.
Just a thought for the day. Probably will not work since these platforms are not know to be supportive of the prosperity of the US population.
Labels:
Pandemic
NJ 2020 03 29
The above is the total as of 1PM today. The growth continues.
However the rate of growth is slowing somewhat consistently.
The total by county are above. Bergen still is the hot spot. Essex, Monmouth, Middlesex, Union, Hudson, Passaic and Ocean are secondary. From Morris down the total are not that extreme.
The above is the % change total by county. The above is the infections per PoP. Bergen is clearly the hot spot. 0.23% of the population is infected. When I see these facts and then hear Fauci I wonder what is the basis for his numbers. NJ is second in total but the infections are in a few hot spots.
This is the % PoP per day change.
Finally the above is the Normalized Daily % change for today by County. One can clearly see the targeting that must be done.
However the rate of growth is slowing somewhat consistently.
The total by county are above. Bergen still is the hot spot. Essex, Monmouth, Middlesex, Union, Hudson, Passaic and Ocean are secondary. From Morris down the total are not that extreme.
The above is the % change total by county. The above is the infections per PoP. Bergen is clearly the hot spot. 0.23% of the population is infected. When I see these facts and then hear Fauci I wonder what is the basis for his numbers. NJ is second in total but the infections are in a few hot spots.
This is the % PoP per day change.
Finally the above is the Normalized Daily % change for today by County. One can clearly see the targeting that must be done.
Labels:
Pandemic
Subscribe to:
Posts (Atom)




