Tuesday, April 28, 2020

my reaction to the modeling update in canada is that they're still over-estimating the mortality rate by about a factor of 10, despite apparently having roughly halved it. i'm not going to repeat myself.

i'd guess around 2-3% of the population is already infected, but that is national, and a bad calculation, conceptually - it would be lower than that in areas like manitoba or saskatchewan, and higher than that in the 401 corridor. as mentioned, off the cuff calculations suggest that the gta could be nearing 20%, even as swift current is still less than 1%.

as the nature of how this disease spreads is inherently local, it's probably a bad idea to rely on federal modeling, and i'm not really going to blame the feds for bad modelling - asking them to do this is actually really the wrong approach. you don't even really want to even look at provincial models, except maybe in small provinces like pei. the six major provinces in this country are each geographically equivalent to the order of 10-15 different states, and the low population density just exacerbates the point. you really want to look specifically at cities, and metropolitan areas.

there are areas of canada that are so remote and so sparsely populated, that they might not have serious epidemics at all and, if they do, may only see first waves in the fall. even a city like thunder bay, which is a substantive population centre, is unlikely to be experiencing much through traffic of the sort that spreads disease.

it's cities like toronto, montreal, calgary and vancouver (as well as border cities like windsor and niagara falls) that are potentially moving into the upswing of the curve over this week and the next, and that you shouldn't be expecting to peak for quite a while, still.

but, the basic analysis is that when you take diverse data sets and combine them too much, you water them down, and just create a worthless mess. that's what the updated federal models are doing - but that's all they ever could do. rather, you need to separate the data very carefully and look at things hyper-locally....

so, if i tell you it's propaganda, you should realize that it's intended to be.