Building Covid 19 tool to help forecast hospital beds

What a frilling week. Market turmoil, oil crashes, and Covid arrived to come kick our asses. I chose to not sit idly by as a spectator, and instead forge a weapon (web app) that will help hospitals around the world predict the wave IN THEIR AREA and tell them how many beds they need and if they will be swamped.

It needs to predict when the wave arrives, and how many of each type of beds will be needed. Standard, ICU, and Critical. Plus, as the data for Covid gets better, the user can change the inputs easily on the webpage.

I haven’t named it officially yet, maybe call it COBE Tool – COvid Bed Estimator Tool. I will put in in Shiny at shiny.io hosted.

Early results show what will happen if Covid19 burns without social distancing. 72% of America catches it.

Tools used : R, with packages Epimdr, Desolve, Shiny. The core code for generating the S E I R table is below.

require(deSolve)
times = seq(0, 450, by=1)
#paras = c(mu = 0, N = 1, beta = 1000, sigma = 365/15, gamma = 365/5)
sigma<- 1/5 # this is     The rate at which an exposed person becomes infective. Inverse of 
gamma<-1/15 #  The rate at which an exposed person becomes resistant (recovers)
paras = c(mu = 0, N = 1, beta = .1173, sigma = sigma, gamma = gamma)
start = c(S=(.999) , E=0, I=.001, R = 0.0)
#start = c(S=0.06, E=0, I=0.001, R = 0.939)
out=ode(y=start, times=times, func=seirmod, parms=paras)
#init       <- c(S = 1-1e-6, I = 1e-6, R = 0.0)
#parameters <- c(beta = 1.4247, gamma = 0.14286)
#out <- ode(y = init, times = times, func = sir, parms = parameters)
out <- as.data.frame(out)
head(out)
tail(out)
plot(out$time,out$R)
plot(out$time,out$I)

Posted in Uncategorized by with comments disabled.

CoVid19 forecasting cumulative cases

I do a bit of competition forecasting. Almanis, Good Judgement Open, and the Hybrid Forecasting Challenge. Some are open, some you must be invited. I’ve found that I’m good at battlefield casualties and epidemiology. Here are my thoughts on CoVid19, and the chances of having greater than 100k cumulative cases by 20MAR20

Finally the Markets are realizing CoVid19 will affect the economy.

Instantaneous forecast from yesterdays increase is 20 days left *1753 new yesterday +85403=~ 120,000 cases.

The instantaneous estimate is no longer trending down, Containment is generally working in China, but CoVid19 has now gone into the wild in other countries. I think there will be no hope of <100k by 20MAR20
But the numbers are increasing more linearly than geometrically.

The real numbers for WHO cases over the last 4 weeks are:

85403 (SitRep Day 40) – not a full week
80239 (SitRep Day 36),
73332 (SitRep Day 29)
43103 (SitRep 22),
20630 (SitRep 15)

Extrapolated weekly estimate is 85403-77794 * 3 weeks left = ~108,000
This shows signs of slowing that the Chinese fight to drop Ro is working. From the model below, I show Ro to be ~1.1 to 1.2 where last week I thought it to be 1.7 roughly.

However, now that is is wild in other countries who aren’t prepared for the Draconian methods China executed. I expect the Ro to begin to rise again.

There is a slim chance to think CoVid-19 will stay below 100k cases as I’m beginning to believe the virus is currently ‘contained’ in China. Also possible is if you believe other countries will systemically under report for political ends. Iran clearly is not reporting real numbers at this time.

Of new concern is the finding previously unknown clusters, even in the USA. Of note is this is a secondary infection not primary from travelers from China. That reinforces my thinking that many second and third world countries have wild clusters established and will begin to spread fast. They may not detect them in time to add to 20MAR WHO report.

I recommend folks follow Dr Campbell on Youtube to get his input, as he has foreign health workers with first hand accounts in Iran emailing him. Those accounts reveal hundreds to thousands infected.
https://www.youtube.com/watch?v=IL7oGRKETZM

Another great resource is https://www.arcgis.com/apps/opsdashboard/index.html#/bda7594740fd40299423467b48e9ecf6 or their mobile friendly version https://www.arcgis.com/apps/opsdashboard/index.html#/85320e2ea5424dfaaa75ae62e5c06e61

The quarantine’s goal is to drop the Ro number from ~2.3 down to less than 1 hopefully, but more likely 1.2 IMHO so we need to adapt that to our numbers.

My assumptions are :
1) too late for containment
2) Wild clusters are undetected in many other countries
3) Iran’s clusters are far worse than feared.
4) The reported cases will increase quasi-geometrically perhaps linearly but not exponential until 20MAR20
5) The new post quarantine Ro is still above 1.1 and below 1.5

Raw data collected via Tencent on the official numbers is https://news.qq.com/zt2020/page/feiyan.htm
yea, its in Simplified Chinese.
-edit- English translate https://translate.google.com/translate?hl=en&sl=auto&tl=en&u=https%3A%2F%2Fnews.qq.com%2Fzt2020%2Fpage%2Ffeiyan.htm

An excellent model that lets you adjust inputs is here:
https://art-bd.shinyapps.io/nCov_control/
(using Ro ~ 1.2 ‘ Effective reproductive number with control’ fits the data best)

I encourage all to play with that model, as it shows the case load with default Ro vs post quarantine Ro and the real cases reported. More on that model here: https://annals.org/aim/fullarticle/2760912/reporting-epidemic-growth-reproduction-numbers-2019-novel-coronavirus-2019-ncov

In all, the infections are showing to grow steadily. If CoVid-19 establishes in South America we may get to 200k quickly, not so sure if it is just the Middle East. The model now shows we’ll hit 100k, but it looks like more likely end at 120k by 20MAR.

I still think we’ll hit 100k cases by March 10.

Probability of <100k by 20MAR20 is 5%. Probability of 100k-200k by 20MAR20 is 92%, Over 200k is 3%


Posted in Uncategorized and tagged , by with comments disabled.

An Atlatl

So I’m an archer not an Atlatl guy. But since it is just a stick that throws a stick, I cut a branch, carved it up, then promptly obsessed about finding good dart material. For Texas, broad leaf ligustrum is the solution. It grows straight after copicing. I have a mott of Ligustrum in my yard so I harvested some 1/2″ poles 5 feet long and made the darts.

Atlatl v1.0

Posted in Uncategorized by with comments disabled.

Alaskan Native American Archery

Having traveled a bit this year thru Alaska, I’ve taken note to photograph all archery equipment I see in the museums. I release these photos under Creative Commons 3.0 license.


Posted in Uncategorized by with comments disabled.

Best Practices creating PAO number system for memory training

So it has been a few weeks since finishing my PAO, many hours later I’ve got it up and running well. Speed is coming along. But now that I’m using it effectively, here are few things I’d wish I’d done differently.

To those coming behind me, who want to build their PAO, here is my advice:

  1. It is normal to struggle to build your 2 digit PAO. I had real trouble finding 100 people who were memorable and I had connection too. In the end when I was about 10 short, I just crammed 10 people that I kinda knew about into the PAO. That was a bad idea, connection seems to be key. You will be using this the rest of your life, take time to build out your PAO people.
  2. Using the Major system to pick the people was just…a waste. Now that I’m using it, I don’t pronounce anything in my head, I just image the PAO. The Major system did nothing for me but handicap the order that I put in people to the PAO. (I even saw someone using Major system for actions and object, which is an aweful idea, adding another burden to remember WHAT the P does and what O it does with with. Go with logical AO for the P) It does nothing for speed or recall. It kinda helped on memorizing it, but that was slow and ultimately overridden but another method, and I’ll address that in #3.
  3. I struggled for days memorizing it. But an hour after I read someone recommended putting it into a dedicated memory palace I did. But I had to make a big palace, 100 loci. I used a video game map, Once I walked thru it, BAM! They they all were, settled into their places within the hour. I’ll address optimizing this in #5.
  4. But I get ahead of myself. The P of PAO are going to be key for speed. I now wish I’d grouped them by decade. 0-10 should have been generics, ie. Astronaut, Doctor, Baker, Butler, Soldier. 10-19 Sports figures, 20-29 Ex girlfriends, 30-39 Star Wars people, 40-49 Harry Potter folks, 50-59 old college professors, etc… This should increase speed in using it. And also make it easier to use 1-40 for Cards, with 4 suits, Clubs, Diamonds, Hearts, Spades (by alpha) all clubs are generics, Diamonds sports figures, ex girlfriends are Hearts, etc. To make it extra fast take care on your memory palace layout.
  5. So when I’m recalling cards now, it is sometimes faster to pull by location. So if you make sure your first 41 locations (remember there is a 00 that goes before 01 where the card deck starts at Ace of Clubs) are grouped with each decade (suit) in a common area. Like 1 big room or 2 rooms with 5. Don’t put 21-23 in the same room with 18-20. Keep the breaks clean, isolate the decades (1-10, 11-20, 21-30, 31-40) cleanly in their own areas. Speed is key on cards and helps with numbers.
  6. Dedicate this memory palace to your PAO. Never reuse it. That will keep it clear of interference.
  7. Be logical on the AO. If you have a P of a Juggler (06), and you are remembering Bill Gates (King of Diamonds) juggling Han Solo’s Blaster, then you quickly remember the ‘A’ is the Juggler 06 and the ‘O’ is Han Solo. Don’t do something weird and have Juggler flying spaceship and Han solo tunneling in minecraft. You will have to do extra cross-tabling recall if you abandon logical actions.

And with that I’m done. I wish someone had given me this advice a month ago.

-mt


Posted in Uncategorized by with comments disabled.

Improved Celeste ripening. Jujube’s has fruit set.

July 4th – All the figs are doing well but the Violette de Bordeaux which is barely growing. At last count, I have 5 fig trees in the front, 4 in the backyard, 2 in pots, and 2 at work.

The Honey Jar Jujube finally set fruit this year, it has been in ground 2 winters. The Shanx Li and the Chico did not set fruit although they bloomed.

Violette de Bordeaux has been a slow grower for 3 years, even with irrigation.

Posted in Uncategorized by with comments disabled.

Roof waterproof coat done with pigment

Put on a pure coat of just portland cement, as I’ve seen it made to a paint consistancy of 1:1 portland to water. It is then painted over cement walls of cisterns to make them waterproof.
In this mix, I put in iron oxide to give it color.

Roof portland waterproof coat put on .


Posted in news by with comments disabled.

Milestone reached! Shed Door Finished and Hung!

Finally, 2 years 2 months later from pouring the foundation, I have finished the door and hung said door on the Shed of Requirement.  This glorious shed I’ve built from foundation up, assembled without power tools, carried stone by stone,  pound for pound 220 feet from the driveway to the site.  The door was supposed to open outwards, but due to geometry restrictions on Gothic doors, it couldn’t open. Not a design consideration I had on my list.  Nonetheless, the door now is hung and opens inwards. If it hits anything, it is a sign I have too much shit in the shed.  With the lifespan of the shed targeted at 200 years, the door is expected to be the first to require replacement in 20-40 years.  Now the shed isn’t done yet, there are many hours left to do.

The inside is a Catalan Vault roof, well, technically a hogbacked oblong dome. I had to attempt the hardest variant. Note, I can lay 7 tile per hour to make that roof.

Shed of Requirement now has a door! Stained cedar door made with hand forged nails cleated thru, and a handle I forged this summer at summer camp.

Stained cedar door I made with hand forged nails cleated thru, and a handle I forged this summer at summer camp.


Posted in news and tagged , by with no comments yet.

The Perch

Archery on posts

Some challenges resonate strongly within us. Sometimes for reasons even we do not know. It is simply part of our baggage we carry on this crazy train we call life. Sunday, at Archery class during summer camp, we were given a challenge to stand on a short pole or a tall pole and shoot as many arrows as possible without falling off…but only allowed to stand on one foot. Typically, people could stand 30-120 seconds, and the best record so far was only 3 arrows. When my time came, I was happy to play the game, as I have great balance. Everyone else had chosen the low perch, screw that, I went high out of pride and arrogance. Once I got up on the high perch, I’m really not sure what happened, it is like I blacked out in a way. I kicked in my Zen and hyper focused. Time dilated, and I had no sense of its passage. My eyes locked on a 2″ square of tree bark and my muscle memory took over, while my brain balanced. I remember my leg beginning to hurt and kicking in my old pain management mindset from 30 years ago. I was planted, that was my pole, I wasn’t leaving. I drew and shot, drew and shot, never looking down, never checking my nock, calling out which target I would hit next. I remember the instructor harassing me, but I was so tuned out and it didn’t get to me. I remember him sticking my arrows in the ground to my right to force me to contort down to the ground to get my next arrow, but I was in Zen and simply retrieved them pluck and grace. I wore that perch as the master and commander. My leg burned but I Zen-ed thru the pain and the searing, I was *not* going to falter, I was *not* going to fall. Apparently, I began to softly talk to myself. Someone thought I was quoting Lord of the Rings. I gave it my all. At some point the instructor refused to give me more arrows. I realized, I wasn’t going to fall off but if I didn’t get down I would collapse off. I voluntarily conceded my perch, and stepped down to Terra firma again. I ceremoniously bowed to the instructor, limped 15 feet away and collapsed down cross-legged. I began to shake and to cry from the intense emotional experience. I’m not sure why, but I believe it was me challenging me, and in the end I was weighed, measured, and found passing. My mettle is known. I was told I’d been on the post 9 minutes, shot 11 arrows, struck the target 10 times. To me, I had done something beautiful and bad-ass, and this time, there were witnesses.


Posted in musings by with comments disabled.

Why you should ignore most questions in the HFC Challenge

I like the HFC competition, for my own reasons. However, most forecasters don’t seem to understand the highly punative scoring system imposed upon them. They will cast a question thinking that because they understand the problem domain more than most, they should answer it. That is flat out wrong, worse, the HFC scoring system inadvertantly encourages casters to participate in more questions because the fewer in a question the more points they can earn. Yet, they don’t seem to process that is a double edge sword.

Lets look at the last few questions that were resoved (correctly, I will point out).
Here are the net points the forecasters got as a group

Paraguayan election -187
Syria strike -303
Cuba election -338
Baffin bay St Lawrence ice extent -305

Let me point out for those who think it is a dash, it is not, it is a negative sign. That is *negative* 187 points. As in, the ‘team Prescience’ would have been better off not forecasting it. At all. But there is an algorithm in the back that will heavily weight the forecasts of the high forecasters and ignore the low ones. So I will avoid discussing the entire team avoiding questions, instead lets look at picking the *right* questions.

If you are wanting to get a high score then lets look at some math. For example the Baffin Bay Ice problem. I got the high score…of 0 points. In fact 13 of us did. The low score got -106. So from a game theory perspective, we risked -106 for achieving 0. That is stupidity.

Another question, the Colorado party Paraguyan question. It started out about 80/20. Top score got only 6 points, the worst got -79. What does that tell you? That you need to be about 13x more confident before you enter a question if it is above 80 or below 20. I could make a heuristic here and say you need to take the percentage ratio 80:20=4 and multiply that by 3 (=12x) to get how much more confident you know more about this question than anybody else. That is, if their is very little variance amongst the casts, so if everyone is in the same range, like 80/20, 82/18, 78/22 then there is low variance.

If high uncertainty reigns. And there are 70/30 and 50/50 and 20/80 and other casts all over the place, then you will be able to win more points as the median estimates will be farther off from your estimate. Like with the Material conflicts in Occupied Palestine question. Almost 1/2 the scores were positive and 1/2 negative, so a favorable question to participate in if you have a small advantage with your judgment.

Conclusions – Triage the new questions carefully. Avoid getting into a question for emotional reasons, or because you know more about it than the average dude at a party.
Avoid questions where there seems to be a consensus in the answer. Unless you really, really, know a shitload more than everyone else *and* you can avoid your own hedgehog bias. You have much to lose and little to gain. On the flip side, if there is chaos amongst the casts, then you can enter the fray with a slight advantage.


Posted in musings and tagged by with 2 comments.