Advertisement
Help Keep Boards Alive. Support us by going ad free today. See here: https://subscriptions.boards.ie/.
https://www.boards.ie/group/1878-subscribers-forum

Private Group for paid up members of Boards.ie. Join the club.
Hi all, please see this major site announcement: https://www.boards.ie/discussion/2058427594/boards-ie-2026

Why we should continue taking actions that are proven to work.

245

Comments

  • #32
    DeVore
    Registered Users, Registered Users 2 Posts: 32,387 ✭✭✭✭DeVore


    Vegeta wrote: »
    Cool, do you guys do mathematical modelling, statistical modelling or both?
    Mostly we do Machine Learning models. I am not sure how statistical models wouldnt be considered mathematical models :):D

    What we do is more in line with statistical models but using things like Neural Networks, Gradient Boost Machines and Random Forests etc.
    Mostly we do it for the insurance industry and its applied against the (vast) amounts of data the insurance companies have about cars and drivers. (ppl, don't @ me about your car insurance, I dont set it and I dont care :D).

    We have some senior people who have worked in virology before but its a different kind of modelling, more "agent based" which we have done some of but which I havent applied here.

    What I'm doing is straight forward "exponential to logistic curve" modelling, hand fitting the data to best first by tuning hyper parameters and then mostly drinking, heavily. :)


  • #33
    Yellow_Fern
    Registered Users, Registered Users 2 Posts: 2,735 ✭✭✭Yellow_Fern


    It is pretty clear that this crisis has been poorly handled. Our gov had to be begged into action by the public and media and clearly they lack the courage to take the strong steps needed to stop the virus stone dead. There are plenty of examples of superior containment and mitigation that we are ignoring.


  • #34
    HeidiHeidi
    Registered Users, Registered Users 2, Paid Member Posts: 32,490 ✭✭✭✭HeidiHeidi


    Yellow_Fern wrote: »
    It is pretty clear that this crisis has been poorly handled. Our gov had to be begged into action by the public and media and clearly they lack the courage to take the strong steps needed to stop the virus stone dead.

    You'll no doubt follow that up with a couple of examples of countries that took those "strong steps", and "stopped the virus stone dead". If you don't mind. :rolleyes:

    (Sorry if that's a derail of the original thread, couldn't resist)


  • #35
    storker
    Registered Users, Registered Users 2 Posts: 5,761 ✭✭✭storker


    Ulysses Gaze wrote: »
    A few people on Boards that think we need to "save the economy at all costs".

    Or are just pretending they think that. Either way, I'd like to see them relocated. To Anthrax Island or maybe Pripyat...just so they can get a sense of what "at all costs" really means.


  • #36
    bloodless_coup
    Banned (with Prison Access) Posts: 1,480 ✭✭✭bloodless_coup


    DeVore wrote: »
    No. I'm Irish. I also started this site 20 years ago. :D

    Then stop talking like an American.


  • Advertisement
  • #37
    DeVore
    Registered Users, Registered Users 2 Posts: 32,387 ✭✭✭✭DeVore


    Nermal wrote: »
    'Might'?

    'Likely'?

    This is the sort of modelling we're doing?
    All modelling is probabilistic. You're trying to predict the future, its by definition unknown.

    As standard the modelling industry will return a prediction ("X% this thing happens") and a confidence rating ("we're Y% sure we're right about that").

    So like we might be 5% sure someone is going to crash their car next year but only 20% confident in that prediction.

    Without lots of hard data about Covid 19 (which is still emerging) we cant have high confidence about long term trends. *HOWEVER* the trend is sooo bad that it doesnt really matter. The variance in the "do nothing" model was from "this is going to be really really terrible" to "We're looking at burning bodies in pyres"... so they took action. (quite correctly imho)


  • #38
    storker
    Registered Users, Registered Users 2 Posts: 5,761 ✭✭✭storker


    DeVore wrote: »
    What I'm doing is straight forward "exponential to logistic curve" modelling, hand fitting the data to best first by tuning hyper parameters and then mostly drinking, heavily. :)

    I think it's disgraceful that you deliberately infected all those innocent little pi's in order to achieve this. I'm emailing PETA about this right now...


  • #39
    CtevenSrowder
    Closed Accounts Posts: 1,909 ✭✭✭CtevenSrowder


    DeVore wrote: »
    "We did a thing, and now the numbers are small We ShOuLd ToTaLly StOp DoInG ThAt ThInG"

    I swear, if I see this sort ...

    Good post, though I've not seen anyone claiming what we are doing now should be stopped just yet!

    Regardless, in your opinion, how long do you think it's feasible to keep the current measures in place? I personally think up until end of May, start of June. I think the 6 months that the UK CMO has said is too long.


  • #40
    Vegeta
    Registered Users, Registered Users 2 Posts: 11,392 ✭✭✭✭Vegeta


    DeVore wrote: »
    Mostly we do Machine Learning models. I am not sure how statistical models wouldnt be considered mathematical models :):D

    I know, I know :D statistics is a field within mathematics but in my limited experience, mathematical modelling has been used to refer to modelling that has its origin in applied mathematics and includes a ton of differential calculus etc. So for stress, strain, heat, brain tissue deformation..... those type of problems
    What I'm doing is straight forward "exponential to logistic curve" modelling, hand fitting the data to best first by tuning hyper parameters and then mostly drinking, heavily. :)

    Recently completed some training in the field so was interested to see what type of problems a dedicated company were working on. What kind of tools are you using for the analysis? Python? Matlab?


  • #41
    CtevenSrowder
    Closed Accounts Posts: 1,909 ✭✭✭CtevenSrowder


    Yellow_Fern wrote: »
    It is pretty clear that this crisis has been poorly handled. Our gov had to be begged into action by the public and media and clearly they lack the courage to take the strong steps needed to stop the virus stone dead. There are plenty of examples of superior containment and mitigation that we are ignoring.

    Christ. How many times does it have to be pointed out that they followed the advice given by the HSE and CMO?


  • Advertisement
  • #42
    Nijmegen
    Registered Users, Registered Users 2 Posts: 3,086 ✭✭✭Nijmegen


    At the bottom of most of the calls for this to be treated in a less severe fashion is fear: Fear of the consequences of the medicine, fear of the disease itself (leading to latching on to anything that evidences it to be less severe than it is) and fear of change.

    I think the only way to deal with this fear is through carefully and plainly explained data, such as the above; and by using calm and reasoned experts such as the ones we have here from the medical community talking to the population daily.

    Good initiative.


  • #43
    FVP3
    Closed Accounts Posts: 1,187 ✭✭✭FVP3


    While what DeVore says is correct, it is hard to see how we can therefore get out of this without antibody testing.

    Politically 2 weeks is about the maximum for a lockdown as much as we can do it here. If we then get a reduction in the rate of increase, then we will probably open up work at least, urging people to WFH where possible, cocooning older people. If the infection rates then start to increase again, we may lockdown again.

    Cant see pubs opening, except perhaps with some outdoor facilities or restricted entry during the inbetween periods.


  • #44
    Wetbench4
    Registered Users, Registered Users 2 Posts: 1,233 ✭✭✭Wetbench4


    I have a question.
    How can we model anything for the future, or even tell whats going on right now, when we know there aren't enough tests? The numbers we see each day aren't accurate at all. The only accurate measurements we have are the numbers in icu and fatalities. We could easily be close to the original estimation of 15000 cases and imo we probably are.


  • #45
    rahmalec
    Registered Users, Registered Users 2 Posts: 193 ✭✭rahmalec


    That's a great video and really well explained, thanks for posting.
    I look forward to watching his other video on mathematical concepts.


  • #46
    Nijmegen
    Registered Users, Registered Users 2 Posts: 3,086 ✭✭✭Nijmegen


    FVP3 wrote: »
    While what DeVore says is correct, it is hard to see how we can therefore get out of this without antibody testing.

    Politically 2 weeks is about the maximum for a lockdown as much as we can do it here. If we then get a reduction in the rate of increase, then we will probably open up work at least, urging people to WFH where possible, cocooning older people. If the infection rates then start to increase again, we may lockdown again.

    Cant see pubs opening, except perhaps with some outdoor facilities or restricted entry during the inbetween periods.

    I was thinking about the slow open with antibody testing scenario. I think it could lead to a lot of social friction and other problems. Bit dystopian to be going around with your government ID to say you've had the virus and are therefore a privileged citizen who can go to the pub.


  • #47
    Jaden
    Registered Users, Registered Users 2 Posts: 2,394 ✭✭✭Jaden


    I have a question: What kind of twisted loon uses the left and right mouse buttons to strafe left and right playing QW?

    Questions on COVID-19 - None. You're right about our actions on combating this pandemic - more of same, please.


  • #48
    FVP3
    Closed Accounts Posts: 1,187 ✭✭✭FVP3


    storker wrote: »
    Yes, because The Economy. :)

    Yes, but while I am not urging an end to these restrictions now, or in the very near future there will come a time when the economy will have to become an issue. This country went into this crisis with high debt level, so as there will be massive layoffs and increases in social protection, there will also be massive increases in debt. Banks might be affected again if the payments of mortgages stops.

    Unless the ECB and other CBs monetize all the debt of all the Euro zone, or some other mechanism, then we will be in dire economic trouble worldwide. Some say that this time there aren't as many underlying problems in the banks, neither are they robust enough to endure liquidity issues for months.

    So there is no clear solution.


  • #49
    PMBC
    Registered Users, Registered Users 2 Posts: 1,123 ✭✭✭PMBC


    I've read a lot of your posts since joining and find you to be very reasonable and rational - all round a good guy. Thanks for this thread and for good scientific information which is what is needed in a time like this.
    The vast majority of people are behaving as required; the numbers, that you averted to, dont lie. Those who arent it is out of lack of knowledge or understanding.
    I will be hit financially by this and am not eligible for any government support because of my age. But we will get through it and we will show how strong we are in the way we do and will continue to help one another.


  • #50
    DeVore
    Registered Users, Registered Users 2 Posts: 32,387 ✭✭✭✭DeVore


    Vegeta wrote: »
    I know, I know :D statistics is a field within mathematics but in my limited experience, mathematical modelling has been used to refer to modelling that has its origin in applied mathematics and includes a ton of differential calculus etc. So for stress, strain, heat, brain tissue deformation..... those type of problems



    Recently completed some training in the field so was interested to see what type of problems a dedicated company were working on. What kind of tools are you using for the analysis? Python? Matlab?
    We use Python, Matlib, SciKitLearn and a few of the other basic libraries. Then we build our own models, usually ensemble methods of several approaches. If you want to work in the industry, learn those but try to learn them "formally" because there are a lot of people knocking about who try to pass themselves off as Data Scientists who really just tinkered with some code and watched a few youtubes.


  • #51
    FVP3
    Closed Accounts Posts: 1,187 ✭✭✭FVP3


    Nijmegen wrote: »
    I was thinking about the slow open with antibody testing scenario. I think it could lead to a lot of social friction and other problems. Bit dystopian to be going around with your government ID to say you've had the virus and are therefore a privileged citizen who can go to the pub.

    Younger people will try picking up the virus in that scenario.


  • Advertisement
  • #52
    johnfás
    Registered Users, Registered Users 2 Posts: 1,344 ✭✭✭johnfás


    CtevenSrowder wrote: »
    Good post, though I've not seen anyone claiming what we are doing now should be stopped just yet!

    Regardless, in your opinion, how long do you think it's feasible to keep the current measures in place? I personally think up until end of May, start of June. I think the 6 months that the UK CMO has said is too long.

    What she said was life may not be back to normal until 6 months. What that might mean is that large concerts / sports events are not licensed for 6 months or that hospital visitation may still be curtailed. She didn’t say that they would have a lockdown for 6 months.


  • #53
    CtevenSrowder
    Closed Accounts Posts: 1,909 ✭✭✭CtevenSrowder


    rahmalec wrote: »
    That's a great video and really well explained, thanks for posting.
    I look forward to watching his other video on mathematical concepts.

    If you are interested in such things, you could check out PBS Infinite series, just search it in YouTube. And PBS Spacetime for the physics version.


  • #54
    nthclare
    Banned (with Prison Access) Posts: 3,312 ✭✭✭nthclare


    I work with plants and sometimes I can tweak germination propagation and flowering times so certain generic hybrids can be almost timed to flower in certain times and temperatures etc comes into play.

    I don't want to share my techniques, results or go into specific details.

    So my background is in botany and horticulture, I know jack sh1t about multicellular eukaryotic's, viruses etc

    But supposedly there's a timer on this virus and I know im getting into conspiracy theories, but what if the virus was designed manipulated and has a life span and it just cuts off at a certain point.

    April 9th seems to be what im hearing, it'll start to fade away from the 6th or 9th of April.

    Forgive my dyslexia and word salad, but hopefully someone has an idea of what I'm on about.


  • #55
    CtevenSrowder
    Closed Accounts Posts: 1,909 ✭✭✭CtevenSrowder


    johnfás wrote: »
    What she said was life may not be back to normal until 6 months. What that might mean is that large concerts / sports events are not licensed for 6 months or that hospital visitation may still be curtailed. She didn’t say that they would have a lockdown for 6 months.

    Fair enough. I never read what she said just a headline so I hold my hands up!


  • #56
    FVP3
    Closed Accounts Posts: 1,187 ✭✭✭FVP3


    So what's easy to continue for 6 months to a year is:

    stopping all sports events and large gatherings, including mass, churches, weddings, events in general. Done anyway.

    Possible is:

    Keeping pubs and restaurants closed fully. Cocooning older people.

    Not possible is:

    What we are doing now.

    If we keep non essential businesses closed for 6 months then we may have nothing to come back to.


  • #57
    DeVore
    Registered Users, Registered Users 2 Posts: 32,387 ✭✭✭✭DeVore


    Wetbench4 wrote: »
    I have a question.
    How can we model anything for the future, or even tell whats going on right now, when we know there aren't enough tests? The numbers we see each day aren't accurate at all. The only accurate measurements we have are the numbers in icu and fatalities. We could easily be close to the original estimation of 15000 cases and imo we probably are.

    Ahh, welcome to my life.

    We NEVER know enough data. In anything.

    Sooo.... we guess. Yep, we guess.

    BUT, we dont just stick our finger in the air and guess. We have ways of "guessing" that are built on solid foundations. Now, without hard data, you are still "guessing" but you can have some confidence in your guesses.


    Imagine you have a graph of data. Now, Paddy collected this data dilligently every day. He recorded it in your database and you come to look at it one day. You graph it and it looks like this:

    Eh, Paddy... why is there a hole in our data?
    Ah well boss, see, I had a feeeed of pints the night before, and didnt come to work that day. I was polluted, it was a mad sessh.

    So, now we have to guess Paddy's missing hangover day but like come on, its pretty obvious. I mean every other day is the same value, well... we can kinda confidently guess that day would have been too. I mean, we cant be 100% sure and we can scowl at Paddy but yeah, theres a damned good chance its the same value that day too.

    Suppose the graph looked like this: 
    \
 \
  \
   
    \
     \

    Well, its a slope now. But we can still "guess" the missing value.

    Thing is, on any graph, if you zoom in real close, its kind of a straight line, if you dont look too hard. So you can do this with a lot of missing values.


    In terms of missing tests for Covid, yeah, there are some huuuge holes in the data. Some of it is questionably accurate and some of it isnt being collected or isnt being released so our "guesses" are pretty big but you also have other countries you can "cog" from and presume (reasonably) that similar things will happen.


  • #58
    redmgar
    Registered Users, Registered Users 2 Posts: 681 ✭✭✭redmgar


    We are not truly 2000, there are 10,000 waiting to be tested based on the new criteria (higher chance of being positive) so you could probably multiply that number.


  • #59
    Yellow_Fern
    Registered Users, Registered Users 2 Posts: 2,735 ✭✭✭Yellow_Fern


    CtevenSrowder wrote: »
    Christ. How many times does it have to be pointed out that they followed the advice given by the HSE and CMO?

    and that is not beyond criticism. Id like to blame the HSE but the buck stops with the gov.


  • #60
    Kivaro
    Closed Accounts Posts: 4,105 ✭✭✭Kivaro


    This is an important subject, and thank you Dev for the informative posts and simulation video.
    The thread title might give people a headache though, and might direct readers away from it.


  • Advertisement
  • #61
    ILoveYourVibes
    Registered Users, Registered Users 2 Posts: 15,242 ✭✭✭✭ILoveYourVibes


    We should understand the thing better keep doing the other thing ..until the thing goes totally away...and we should learn more about the thing and also learn more about how we can make lives better for people doing the thing.

    Also the things no are not low ....we just ran out of things ..and tests ...and our testing is like a few days old by the time we get the results.

    Death to the thing. **** the thing.


Advertisement