Profile picture
, 3 tweets, 1 min read
My Authors
Read all threads
Working on something where you can model the number of *detected* cases of a disease as a function of the number of *actual* cases and various assumptions about how/how many tests are conducted.
This was supposed to be a simple model (simple enough to do in EXCEL) but it's clear that you can get some extremely funky behavior where actual and detected cases don't line up very well depending on what assumptions you make about testing.
If you're one of the epidemiologists I follow, send me a DM with your e-mail address and I will send you a copy. Will want your feedback though!
Missing some Tweet in this thread? You can try to force a refresh.

Enjoying this thread?

Keep Current with Nate Silver

Profile picture

Stay in touch and get notified when new unrolls are available from this author!

Read all threads

This Thread may be Removed Anytime!

Twitter may remove this content at anytime, convert it as a PDF, save and print for later use!

Try unrolling a thread yourself!

how to unroll video

1) Follow Thread Reader App on Twitter so you can easily mention us!

2) Go to a Twitter thread (series of Tweets by the same owner) and mention us with a keyword "unroll" @threadreaderapp unroll

You can practice here first or read more on our help page!

More from @NateSilver538 see all

Profile picture
Nate Silver
@NateSilver538
Newly-reported US COVID-19 deaths:

Mon 4/6—1182
Sun 4/5—1184
Sat 4/4—1352
Fri 4/3—1178
Thu 4/2—1084
Wed 4/1—954
Tue 3/31—807
Mon 3/30—511
Sun 3/29—463
Sat 3/28—435
Fri 3/27—367
Thu 3/26—263
Wed 3/25—225
Tue 3/24—204
Mon 3/23—73
Sun 3/22—126
Sat 3/21—53
Fri 3/20—59
Thu 3/19—48
So, there's a lot going on here.

There's been a slowdown. But also important: there may be some day-of-week effects. The last two Tuesdays were quite bad. Possible there's a backlog of cases on weekends that then finally make it into Tuesday's reporting.
Source as usual is @COVID19Tracking. Also, there's *clearly* been a New York slowdown, which is tending to dominate the data; NY vs. rest of country may be on somewhat different trajectories.

Read 5 tweets
Profile picture
Nate Silver
@NateSilver538
The media narrative was **extremely terrible** for Biden when he came back to win South Carolina by 30 points after having been tied in the polls days earlier.
That SC win created a lot of momentum/favorable media heading into Super Tuesday, and Super Tuesday created further momentum. But Biden really earned that SC landslide by having the trust of black voters and non-college whites who are key in SC (+ Clyburn's endorsement).
Only two candidates (Biden and Bernie) ever really had a mutli-racial, multi-class coalition, which is usually how win the D primary. Of those 2, only Biden really built bridges to the rest of the party, which is usually how you win *any* party nomination, Trump an exception.
Read 5 tweets
Profile picture
Nate Silver
@NateSilver538
Another issue with charts drawn on a log scale is that they literally compress what has been happening *recently* when the stakes are very high (10s of thousands) while placing a lot of visual emphasis on what the slope was when you were going from say 10 cases or deaths to 100.
The US's slope has flattened quite a bit in recent days, especially in New York. Not necessarily flattened any more or less than you'd expect. But it has flattened. And it's quite hard to pick that up from those charts.
To put this in a super dorky way, the shape that an epidemic curve takes is not actually exponential but sigmoidal (S-shaped / it levels off at some point) and putting a sigmoid on a log scale makes it harder to see when you hit the inflection point and start leveling off.
Read 3 tweets

Related threads

Profile picture
Seth Abramson (@🏠)
@SethAbramson
On the first day of tracking whether the new coronavirus model Dr. Birx is using is accurate, the data on the ground exceeded the highest end of the model for deaths and new cases by 15%+.

Assumptions re: what will be possible in August must be rethought. cnn.com/2020/04/02/pol…
PS/ In other words, far too many big decisions are being made that are predicated on the assumption that by August daily coronavirus deaths will have dropped to zero and Americans will be psychologically prepared to return to normal life. That assumption is wrong in several ways.
PS2/ First, so far nothing is happening consistent with a model that says daily deaths will be at zero in August, and even a daily death rate in the hundreds would be consistent with the rate America had when many states put themselves in lockdown. Why would August be different?
Read 8 tweets
Profile picture
Rohit Ramchandani
@DrRramchandani
Growth rate is key during a disease outbreak. The rate of growth in the the total confirmed #cases in #Canada seems to be showing signs of slowing and pulling away from the continued steep trend line seen, for example, in the United States (Green). #COVID19 #coronavirus 1/25
This is a +ve sign, but still very early to tell, esp. with an est. 1M+ #Canadians having just returned from travel, and #COVID-19 #testing only targeting >severe cases (due to rationing & backlog in lab testing), we don’t have a good grasp on the level of community spread. 2/25
The number of total cases is what we want to know, but the true number of these is not known. Not by us epidemiologists, nor by any other research, governmental or reporting institution. 3/25
Read 25 tweets
Profile picture
Ryan Seamus McGee
@RS_McGee
We would like to share an extended SEIR model and easy to use python package for studying population structure, social distancing, testing, tracing, and quarantining—including stochastic implementations of these models on dynamic networks. (thread)
Here’s a brief overview, but the model and code are fully documented in the GitHub README. This work is a collaboration with @ocornejopopgen, @evokerr, @mtanaka000, and @CT_Bergstrom.
We use an extended SEIR model with compartments for detected cases. Individuals are tested at some rate, and positive tests move to a detected infection state. Known cases may then have different parameters or connectivity representing targeted interventions or treatment.
Read 14 tweets
Profile picture
Dr Gaetan Burgio, MD, PhD.
@GaetanBurgio
I wanted to get my head around what is happening in Australia with #COVID19 and how appropriate our response is. First the number of detected cases in Australia. We are well and truly in exponential state (today +281) with a growth rate of +20% per day. doubling time 3.5 days.
While the number of detected cases are climbing rapidly, I wanted to see how it compares to other countries. So I compared the number of #COVID19 detected cases to various countries (adjusted by population size) and indeed we should be worried.
An explanation is 1/ We are good at detecting cases in the community or from travelers or 2/ Many cases are mild or not severe. So I looked at the death rate (adjusted by population size). We are doing similarly to Germany for now. Look at Italy, this is truly frightening
Read 9 tweets
Profile picture
Benedict Evans
@benedictevans
Alexa is a huge success. Amazon invented a category and has the AI leader Google and the device leader Apple in its wake. 100m units sold! (So maybe 50m in use). But... That’s a pretty contingent success. What actual benefit does Amazon get from this? 1/
Amazon isn’t a reach company, like Google or FB, and it’s not making margin from the devices, unlike Apple. It needs to sell things. How does Alexa do that if people only use if for trivia questions, music and kitchen timers? There needs to be a next step 2/
That is, one could say that Alexa has achieved some level of product-market fit for consumers, but not, yet, for Amazon. 3/
Read 7 tweets

Trending hashtags

#province #ISWVBingo #Coronaindia #s #Boulder #HOWEMBARRASSING #Nebraska #Topic #blackfacelawyer #keralafightsCorona #Namibia #CMA #StayTheFHome #Canada #COVID2019 #Threadnought #PCR #COVID19 #CoronavirusOutbreakindia #currency #BOOM #covidindia #QTNA #COVID19Bangalore #pandemic

Embed code for your website

×
Follow Us on Twitter!

Did Thread Reader help you today?

Support us! We are indie developers!

This site is made by just three indie developers on a laptop doing marketing, support and development! Read more about the story.

Become a Premium Member ($3.00/month or $30.00/year) and get exclusive features!

Become Premium

Too expensive? Make a small donation by buying us coffee ($5) or help with server cost ($10)

Donate via Paypal Become our Patreon

Thank you for your support!