Where is Priti Patel from? Just wonderin’ about her accent as she seems unable to pronounce any word ending in “-ing” without dropping the “g”. Is this a regional thing I’ve not encountered yet?

Beth Rigby from Sky also does this. Gets right on my wick. You’d think that someone who waffles on, for a living, could address this?

I don’t mind regional accents at all. Just not familiar with this variation.

The slides the government show for hospital admissions – is this total number in hospital that day or new admissions each day?

Is this a regional thing I’ve not encountered yet?

Sadiq Khan does the same.

I am familiar with the Black Swan, yes. Most risk assessment make assumptions about distributions. Then they add assumption to assumption to… that gives you an idea.

Now if you assume bell shaped curves, that works ok, and mathematically is relatively simple. BUT surprisingly, life in the extremes of the tails is NOT bell-shaped, so rare events may or may not be quite as rare as you think. See 100yr floods. The challenge is guessing what distributions look like, and how to validate those assumptions.

[TL:DR] models are good but data is always more variable. Identifying that early is a Good Thing.

Caitlin Moran writes an article in todays Times saying how easy lockdown is and gets outed for attending a party yesterday by the nextdoor neighbour.

Where is Priti Patel from? Just wonderin’ about her accent as she seems unable to pronounce any word ending in “-ing” without dropping the “g”. Is this a regional thing I’ve not encountered yet?

Radio 2.

I think we are going to hit the biggest problem with social distancing now, the novelty has worn off people are getting bored, the weather is not helping.
Heard a few bikes about today aswell.
Plus some are stuiply clutching to 3-4 weeks expectation. Wuhan needed 12 weeks lockdown to think we need anything else is a bit naive.
Prepare for the worst hope for the best.

She’s an MP for Essex, innit.

Radio 2.

🙂

Might be true for all I now. I never listen to it

Damn wasted a joke.

Radio 2

Just home officin’?

You can play “pin the tail of the epidemic on the curve” game. There are two models going on there, is a near term time-series (wiggly) model and a long term (epidemic) prediction, when they converge, then the final figures will be more accurate. I have these for EVERY country in the global dataset. This is just the UK. Uses today’s data I just downloaded from ECDC. Hope that makes it a bit more real.

Patel refusing to apologise for the lack of PPE equipment available to front line workers.

There are two models going on there, is a near term time-series (wiggly) model and a long term (epidemic) prediction

What is the “tuning parameter” that would make the epidemiological model fit the data better?

Long-term prediction is going to be sensitive to parameters estimated from the global epidemic model, and then “selected” specifically for the UK data (posterior), so I ask the epidemiological model to provide a 7-day prediction as well as final size. There is no tuning per-se, it is a quality control as to whether the epidemiological model can give a near-cast. At the moment, in the UK it cannot,so I rely on the time-series model instead. For Italy and the US, they have converged perfectly. Spain is also there. BTW I don’t fit incidence, I fit cumulative cases and deaths, then calculate the daily rates – that gives the classic curves. Current error is 11% for the time series and 15% for the epidemiological models. There are 50 countries in today’s dataset

Quick, deploy the non-apology!

“I’m sorry if people feel that there have been failings.”

After being asked twice if she would apologise to NHS staff and their families over the lack of “necessary PPE” that has been linked to NHS workers becoming infected and dying, she said:

“I’ve been very clear in what I have said and I’m sorry that people feel that way”

So – probably getting the wrong end of the stick, but to rephrase – what are the parameters that control the shape of the epidemiological model, and what would have to be adjusted to make it have a bigger peak?

Quick, deploy the non-apology!

“I’m sorry if people are upset because staff/family have died because we didn’t preload the NHS and care system with extra PPE in preparation for the epidemic that seemed unavoidable months ago. But you could just be more like me, shrug it off, assume staff will just do their job without adequate protection, because they care even if I seemingly do not.’

NHS staff as pawns in a game of chicken with a poorly understood virus. Are we allowed to be **** angry? Or should we just nod?

I made my own beer garden. Hope those of you with the outside room did to. Too few people in it though, will be glad to be out of isolation as soon as possible… beers were cheaper, and beer gardens don’t normally have armchairs… but still not as good as the real thing… you need people for that.

DrJ the model has three parameters, growth rate, carrying capacity (end size) and initial starting condition. Each of these have a global mean and distributions to describe country-to-country variability. The height of the peak is informed by the global curve and the posterior is estimated for each country with an assumption about country-to-country variability. Peak time is informed by starting condition as rate is largely fixed in the model.

Of note, but if I just model the U.K. data then you cannot estimate anything from the U.K. data! You have to have passed the peak as inflections provide the most parameter information. A global model means any country ahead informs on those behind.

@TiRed thanks – just trying to understand better 🙂

Is it realistic to assume that the UK is still running a fortnight behind Spain and Italy?

“Three hundred thousand and thirty four, nine hundred and seventy four thousand tests carried out.”pic.twitter.com/JoYVCP5GQz

— James Manning-Monro (@JamesManning) April 11, 2020

“Three hundred thousand and thirty four, nine hundred and seventy four thousand tests carried out. How many?

Bruneep,

Where’s Diane Abbot when you need her.

Is it realistic to assume that the UK is still running a fortnight behind Spain and Italy?

Almost. Two weeks behind Italy and a week behind Spain. Roughly. We should be over the peak and solidly declining in two weeks time. How long we bump along on the peak is debatable but I’ve always said it was identified to within two weeks. Rate of decline will depend on efficiency of lock down.

The govt don’t seem to be able to share plots other than the most rudimentary data and never on a log scale! I’ve not seen a projection from them yet. BTW, they’ve seen mine.

kelvin
Subscriber

Quick, deploy the non-apology!

“I’m sorry if people are upset because staff/family have died because we didn’t preload the NHS and care system with extra PPE in preparation for the epidemic that seemed unavoidable months ago. But you could just be more like me, shrug it off, assume staff will just do their job without adequate protection, because they care even if I seemingly do not.’

NHS staff as pawns in a game of chicken with a poorly understood virus. Are we allowed to be **** angry? Or should we just nod?

It’s pretty much that from family & friends who work in the NHS.
They feel like cannon fodder, but despite this they have an underlying sense of duty to carry on as they are decent people.

I’m sorry you all think i should apologise for not caring more of you will die than should, but I’m not going to.

You know, I really don’t warm to her.

I hope your numbers are right for us @TiRed.

Was PPE actually available to stockpile in January when the government should have twigged? Every country in the world was after it, and China was shutting down.

We should have had PPE before this even started.

Where is Priti Patel from? Just wonderin’ about her accent as she seems unable to pronounce any word ending in “-ing” without dropping the “g”. Is this a regional thing I’ve not encountered yet?

I can’t believe a man from Aviemore actually wrote this 🙂

TiRed – if I understand the UK graph correctly then we are roughly 20% higher on cases per day than predicted and a bit more on deaths? If so does this point to a serious lack of testing making the cases figure possibly suppressed and our curve projection a bit imperfect?

I’m of the opinion that as we don’t have mass testing up to speed yet that we don’t really know how many cases there are. This makes it very hard indeed to gauge whether the lockdown is working or not and also whether the NHS systems are better or worse than predicted for preventing cases turning into deaths. Please feel free to correct my thinking as I’ve been following your explanations as much as I can throughout but as maths was never my strong point I can sometimes end up going down the wrong rabbit hole!

Turns out those 3.5million test kits Hancock ordered were ‘sold as seen’

https://www.theguardian.com/world/2020/apr/11/reveal-cost-of-35m-unusable-covid-19-tests-health-chiefs-told

So after 3 weeks of hiding Patel numberwanged the press conference, I suppose she’ll not be back, who’s up next, 1000 yard state Raab, many homes Jenrick, no information Sharma, blown it that just leaves signal , Gove & captain BS-Hancock . I bet they can’t wait for Johnson to return!

A lump of dog shite on a stick would give a more reasoned/informative/factual press conference than any of these **** muppets that we’ve had the misfortune to suffer so far

On the topic of model predictions:

It’s a simple SEIR model with breakpoint in R on the date of lockdown. Seems to describe what’s going on pretty well (in a range of places I’ve tried, including Hubei, Lombardy, and various European countries). Rather better than the MRC/IC predictions in fact.

I suppose she’ll not be back, who’s up next

I finally got my important letter from the government. I was suspicious rather than reminding me of the state of affairs a couple weeks back it would actually be a note asking me to confirm when I could do the conference.
In some ways I feel sorry for her. She couldnt exactly turn round and go “look I was trying to make all the foreign NHS staff leave the country. It was Hancock who was supposed to be sorting out the PPE and failed. Go and speak with him”.

somafunk
Subscriber

A lump of dog shite on a stick would give a more reasoned/informative/factual press conference than any of these **** muppets that we’ve had the misfortune to suffer so far

Beautifully put.
I’m stealing that for use elsewhere.

that we don’t really know how many cases there are. This makes it very hard indeed to gauge whether the lockdown is working or not

The model uses a parametric form that is not compatible with conventional mass action. The fact that this is a statistically much better description of the observations (in all countries) provides the evidence of lockdown efficiency.

The other prediction is precisely all that is challenging with time series forecasting. Predictions for the future are based on predictions which have uncertainty. That error is on a log scale. My method for near-casting is quite different and novel. Prediction error is about 15-20%. My parametric model has 15% interval because the dataset is global, not limited to one country.

I’ve done the same for the U.K. 164 regions. There is very strong support that a logistic equation (SIR model) is much poorer than other models.