tag:blogger.com,1999:blog-9959776.comments2021-09-25T09:53:16.694+01:00James' Empty BlogJames Annanhttp://www.blogger.com/profile/04318741813895533700noreply@blogger.comBlogger12213125tag:blogger.com,1999:blog-9959776.post-33657764095941685272021-09-20T15:40:21.221+01:002021-09-20T15:40:21.221+01:00Thanks! I ended up just coding up a google sheets ...Thanks! I ended up just coding up a google sheets custom function to estimate the inverse. Three iterations of Newton's method is more than enough to get within 0.1 km/h. I've attached my function here in case it's of interest to anyone else!<br /><br /><br />const a = 1.86e-02;<br />const b = -5.37e-04;<br />const c = 2.23e-05;<br />const d = 1.33e-05;<br /><br />function SPEEDTOPOWER(speed, weight, height) {<br /> return a * weight * speed + (b + c * weight + d * height) * speed * speed * speed;<br />}<br /><br />function POWERTOSPEED(power, weight, height) {<br /> let speed = 40; // Initial guess - about right for TTT<br /> const A = a * weight;<br /> const B = b + c * weight + d * height;<br /> // Use Newton's method to find a root of f = A * speed + B * speed^3 - P<br /> for (i = 0; i < 3; ++i) {<br /> speed = speed - (A * speed + B * speed * speed * speed - power) / (A + 3 * B * speed * speed)<br /> }<br /> return speed;<br />}<br />Simonhttps://www.blogger.com/profile/08252808752314324483noreply@blogger.comtag:blogger.com,1999:blog-9959776.post-267465955313497712021-09-20T10:38:46.983+01:002021-09-20T10:38:46.983+01:00Thanks for comments. No direct inverse, though you...Thanks for comments. No direct inverse, though you could approximate quite well in various ways. If you are trying to give pull targets then just using 2.5W/kg and 1W/cm (as deviations from the team mean) is probably good enough really. The fit without the linear term is a lot worse.<br /><br />I just put in the speed and adjust it until the pulls look about right for the team members (who have been vary a lot from week to week). Seems to work pretty well.James Annanhttps://www.blogger.com/profile/04318741813895533700noreply@blogger.comtag:blogger.com,1999:blog-9959776.post-80078640517736578212021-09-20T03:56:47.379+01:002021-09-20T03:56:47.379+01:00For my TTT simulator, I want to compute speed (on ...For my TTT simulator, I want to compute speed (on the flats) given a specific target power so I really want the inverse of your equation. I think I can just use a couple of iterations of Newton's method with a starting v in the right ballpark (e.g. 40km/h for a TT), i.e. a couple of repetitions of:<br /><br />v1 = v0 - (awv + (-b + cw + dh) v^3 - P) / (aq + 3(-b + cw + dh) v^2),<br /><br />where a, b, c, and d are the 4 constants in your expression.<br /><br />But a closed form expression would be even better and, given that your initial thought was that the power was a function of v^3, which would have a simple inverse, did you try the fit without that wv term? Any thoughts welcome. Thanks again!Simonhttps://www.blogger.com/profile/08252808752314324483noreply@blogger.comtag:blogger.com,1999:blog-9959776.post-71057910837781683342021-09-18T22:27:12.564+01:002021-09-18T22:27:12.564+01:00I was just fiddling around with my TTT spreadsheet...I was just fiddling around with my TTT spreadsheet and thinking that someone out there *must* have worked out the relationship between power and speed on the flat. And here it is! Many thanks.Simonhttps://www.blogger.com/profile/08252808752314324483noreply@blogger.comtag:blogger.com,1999:blog-9959776.post-1099641281370472052021-04-21T18:12:28.163+01:002021-04-21T18:12:28.163+01:00Ah well done that’s a good control set up! ZP data...Ah well done that’s a good control set up! ZP data probably more useful for climbing then, when speeds low enough to ignore draft. For example I used it recently to figure out how what w/kg to hold when aiming for a sub hour Alpe. Tim Andrewshttps://www.blogger.com/profile/12452454527875723780noreply@blogger.comtag:blogger.com,1999:blog-9959776.post-28824645815548956182021-04-21T17:45:30.742+01:002021-04-21T17:45:30.742+01:00Yes ZP is an interesting source for population-wid...Yes ZP is an interesting source for population-wide stats but as you say the draft issue is a big one. Just looking at TT events could work but even then temporal changes in power will affect the results. And I want the calculation calibrated for normal bikes (but no draft, as in the front rider of a TTT team).<br /><br />I did all my runs solo by setting up a private meet-up that the other person didn't attend - I was the only bike on screen. Precisely for this reason :-)James Annanhttps://www.blogger.com/profile/04318741813895533700noreply@blogger.comtag:blogger.com,1999:blog-9959776.post-53707694899345789302021-04-21T14:59:26.738+01:002021-04-21T14:59:26.738+01:00Hi James,
Are you on ZwiftPower? Loads of data on...Hi James,<br /><br />Are you on ZwiftPower? Loads of data on there. See https://www.zwiftpower.com/segments.php?id=10 for the volcano circuit forward (which is what I presume what you did?). It has a population average of:<br /><br />Time: 06:06<br />Power: 252w<br />Weight: 75.6kg<br />Watts/kg: 3.3wkg<br />Speed: 40.1kmh<br /><br />Which seems to fit, but you can filter by weight, power etc. or even plot the time versus w/kg. Though I guess you'd never know if they rode in a bunch or solo, which would make a very big difference. In fact you need to do your tests again with a TT bike rather than AeroRoad to ensure there was no drafting from anyone else on screen... ;-)<br /><br />Cheers,<br />Tim.Tim Andrewshttps://www.blogger.com/profile/12452454527875723780noreply@blogger.comtag:blogger.com,1999:blog-9959776.post-606452034317181952021-04-15T08:17:24.195+01:002021-04-15T08:17:24.195+01:00It seems like you didn't need to mention j was...It seems like you didn't need to mention j was the bottom...William M. Connolleyhttps://www.blogger.com/profile/05836299130680534926noreply@blogger.comtag:blogger.com,1999:blog-9959776.post-20601176715035788122021-04-14T21:46:32.189+01:002021-04-14T21:46:32.189+01:00I am the bottom line! hurrah!I am the bottom line! hurrah!juleshttps://www.blogger.com/profile/02591920483149775255noreply@blogger.comtag:blogger.com,1999:blog-9959776.post-24191324489743254982021-03-23T15:23:23.488+00:002021-03-23T15:23:23.488+00:00Well...the SPI-M-O membership overlaps with SAGE s...Well...the SPI-M-O membership overlaps with SAGE so I think we can be 100% certain that some SAGE members knew of the new doubling time on the 20th. But SAGE did not meet as a committee and therefore could not have taken a collective view still less presented it to ministers. "This information was only considered by SAGE on the 23rd March."<br /><br />It is a bit puzzling as to why SAGE didn't meet over the weekend given the extreme urgency of the situation that SPI-M-O had summarised on the Friday. Perhaps there was some acknowledgment that the govt had to get its ducks in a row in order to be able to respond promptly, so they sat on their hands for the weekend while govt departments sorted out the details. But this is pure speculation. Maybe Vallance was off on a stag weekend.<br /><br />I should also say I heard rumours of an imminent London lockdown a few days earlier, it turns out that this was considered by SAGE/SPI-M and there's even a specific briefing note about it. Just because a rumoured event subsequently came true does not mean that it had been decided at the time that the rumour started.James Annanhttps://www.blogger.com/profile/04318741813895533700noreply@blogger.comtag:blogger.com,1999:blog-9959776.post-51176186572617380622021-03-23T14:05:20.123+00:002021-03-23T14:05:20.123+00:00Not disputing your version of events. But I knew l...Not disputing your version of events. But I knew lockdown was coming to be announced on 23rd (even if I wasn't clear if it would take place immediately or on 24th or 25th) certainly by Sat 21st and I think probably by Friday. This was via a sister living abroad with a contact at an embassy. <br /><br />If the information was spreading like that, then it seems a bit silly to say "This information only reached SAGE on the 23rd March". I am sure you mean and we should interpret that more as<br /><br />'This information only formally reached SAGE minutes on the 23rd March but no doubt informally reached at least some SAGE members on 20th and maybe late 18th and 19th cannot be ruled out but early on 18th can be ruled out.' <br /><br />Anonymoushttps://www.blogger.com/profile/15105620177689423189noreply@blogger.comtag:blogger.com,1999:blog-9959776.post-38167720626319168812021-02-08T08:06:02.174+00:002021-02-08T08:06:02.174+00:00I believe that you assume too small mortality from...I believe that you assume too small mortality from covid. My colleague just completed the analysis on the UK data, and his estimate is roughly 1 death out of 150 infected. This heavily depends on the distribution of infections between age groups, but this is roughly in the middle. Seventy five thousands deaths at January 1st would translate to 11 mln infections, not 15 mln that you assume. Witoldhttps://www.blogger.com/profile/17160611381775852928noreply@blogger.comtag:blogger.com,1999:blog-9959776.post-13577740017980610652021-01-26T14:16:50.485+00:002021-01-26T14:16:50.485+00:00Well I think there is some sort of point, in that ...Well I think there is some sort of point, in that the effect of interventions is noticed in a change in the growth rate rather than in the number of cases directly. However I don't see where they get this two week delay from as cases will respond more quickly than that to behavioural changes. I also haven't looked at the USA data.James Annanhttps://www.blogger.com/profile/04318741813895533700noreply@blogger.comtag:blogger.com,1999:blog-9959776.post-52996389063515722192021-01-26T03:20:39.079+00:002021-01-26T03:20:39.079+00:00Elsewhere, I saw this comment on when effect of ho...Elsewhere, I saw this comment on when effect of holiday contacts would show in US data. (I noted that it had peaked on Jan 8).<br /><br /><i>"Daily cases are irrelevant. The supposed holiday effect would impact transmission. That means that it would affect the *slope* of the logarithm of daily cases, with a delay of a couple weeks. Looking at the logarithmic plot for the USA, it looks like it was steepest around the end of October, so transmission peaked in mid-October. It finally went negative a couple weeks ago, just about when it should have been spiking, according to the holiday transmission theory."</i><br /><br />I really dont get what is meant. Does this make any sense?PhilScaddenhttps://www.blogger.com/profile/05937238628676275303noreply@blogger.comtag:blogger.com,1999:blog-9959776.post-27223669146815442212020-12-22T19:30:45.436+00:002020-12-22T19:30:45.436+00:00I'm sorry to hear that about the vaccination i...I'm sorry to hear that about the vaccination issues. I do know someone was called up as a last-minute replacement for an early slot, and expect transport is a challenge round here. But as long as it's getting used up without too much wastage, it all counts. The first 5 million makes a really big dent in the deaths (and presumably, overall healthcare demands).<br /><br />I'm a bit sceptical of the tone of that article - if "the" vaccine was all sorted in a few days, why are there so many different ones? I do agree that it may be streamlined in future though especially after the current experience.James Annanhttps://www.blogger.com/profile/04318741813895533700noreply@blogger.comtag:blogger.com,1999:blog-9959776.post-70321333574243311642020-12-22T18:43:39.892+00:002020-12-22T18:43:39.892+00:00> the vaccination seems to be progressing smoot...> the vaccination seems to be progressing smoothly and at a decent pace<br /><br />That (to somewhat hijack the discussion, sorry) seems to me somewhat doubtful. Both in terms of the numbers per week so far, and anecdote (from GP: lots of potential slots empty as the system fails to organise enough customers). I do also wonder (see-also <a href="https://nymag.com/intelligencer/2020/12/moderna-covid-19-vaccine-design.html" rel="nofollow">this</a>) if we may come to regret not pushing out the vaccine earlier.William M. Connolleyhttps://www.blogger.com/profile/05836299130680534926noreply@blogger.comtag:blogger.com,1999:blog-9959776.post-57568624884261395862020-11-09T15:17:45.279+00:002020-11-09T15:17:45.279+00:00That's certainly on the to-do list, but I have...That's certainly on the to-do list, but I have to deal with a bit more paid work first (it's more the deadline than the pay). I assume it will work similarly well to the usual covid modelling, which is pretty reliable. In fact I have some plots of that I did just recently for the Lancaster talk, maybe I will post those too.James Annanhttps://www.blogger.com/profile/04318741813895533700noreply@blogger.comtag:blogger.com,1999:blog-9959776.post-23523664222905266242020-11-09T14:18:37.751+00:002020-11-09T14:18:37.751+00:00Now you've put some results up (https://twitte...Now you've put some results up (https://twitter.com/jamesannan/status/1325717106067988480?s=20) I thought I'd ask the question I wanted to put, but will put here cos Twitter is annoying: how would that look in "forecast" mode? Suppose you start running the model with data up to July, then Aug, and so on?William M. Connolleyhttps://www.blogger.com/profile/05836299130680534926noreply@blogger.comtag:blogger.com,1999:blog-9959776.post-60435312427857070152020-11-09T09:23:09.573+00:002020-11-09T09:23:09.573+00:00Hello. Blogwalking & followed here :) Have a n...Hello. Blogwalking & followed here :) Have a nice day!ASTravellerhttps://www.blogger.com/profile/03681661700086089674noreply@blogger.comtag:blogger.com,1999:blog-9959776.post-2953761277658933402020-10-11T23:52:48.115+01:002020-10-11T23:52:48.115+01:00Hi James, thanks for all your very interesting twe...Hi James, thanks for all your very interesting tweets. I wonder what you thought of the KCL app? Their estimated number of daily new cases has been increasing more slowly for a couple of weeks now (10-15% increase over that time (today about 22k)). That would suggest we may be nearing a plateau or peak, but that we might see 150-200 deaths in a few days' time and that this will continue for several weeks? hope you and Jules are both well!CJBhttps://www.blogger.com/profile/09421410661247766429noreply@blogger.comtag:blogger.com,1999:blog-9959776.post-1554191191652859182020-09-21T22:26:22.163+01:002020-09-21T22:26:22.163+01:00This comment has been removed by the author.Graemehttps://www.blogger.com/profile/11007306140530173428noreply@blogger.comtag:blogger.com,1999:blog-9959776.post-39831900601225243602020-09-21T22:25:18.392+01:002020-09-21T22:25:18.392+01:00This comment has been removed by the author.Graemehttps://www.blogger.com/profile/11007306140530173428noreply@blogger.comtag:blogger.com,1999:blog-9959776.post-40205288143400377442020-09-14T11:24:05.675+01:002020-09-14T11:24:05.675+01:00Nick, that's because the 0.75% death rate in t...Nick, that's because the 0.75% death rate in the model is too high currently. Especially with the change to 28d deaths (which undercounts the true number!), there simply aren't enough of them for the cases we are getting. So the model is compromising between the two. I have thought of putting in a death rate that reduces over time but haven't got round to it. It would involve more subjective choices, degrees of freedom...and work :-)<br /><br />Warren, well it certainly seems quite high to me (though not overwhelming) and rising exponentially to me....though the numbers in hospital are still moderate so I guess that is their excuse for sitting on their hands.James Annanhttps://www.blogger.com/profile/04318741813895533700noreply@blogger.comtag:blogger.com,1999:blog-9959776.post-45912114241656656602020-09-14T10:42:07.987+01:002020-09-14T10:42:07.987+01:00Since the beginning of July, essentially every red...Since the beginning of July, essentially every red circle (case count) falls above the magenta line (model median). Is that reasonable, or a bug? It feels as if your model is running cold and not correcting adequately for some reason. Similarly on the deaths chart, almost every data point is below the model median.Nick Barneshttps://www.blogger.com/profile/00057838251997644583noreply@blogger.comtag:blogger.com,1999:blog-9959776.post-14601994706461663352020-09-14T10:13:48.685+01:002020-09-14T10:13:48.685+01:00Good summary - thanks! I am wondering what all th...Good summary - thanks! I am wondering what all this means in policy terms. The Covid Alert Level is currently at Level 3 (general circulation). Level 4 is "transmission is high or rising exponentially", so even on SAGE's estimate we should me moving to 4 (although I admit my understanding of 'rising exponentially' is fuzzy)? https://www.gov.uk/government/news/update-from-the-uk-chief-medical-officers-on-the-uk-alert-level Warren Pearcehttps://www.blogger.com/profile/00262948477766071651noreply@blogger.com