[Nature on] simulations driving the world’s response to COVID-19

Nature of 02 April 2020 has a special section on simulation methods used to assess and predict the pandemic evolution. Calling for caution as the models used therein, like the standard ODE S(E)IR models, which rely on assumptions on the spread of the data and very rarely on data, especially in the early stages of the pandemic. One epidemiologist is quote stating “We’re building simplified representations of reality” but this is not dire enough, as “simplified” evokes “less precise” rather than “possibly grossly misleading”. (The graph above is unrelated to the Nature cover and appears to me as particularly appalling in mixing different types of data, time-scale, population at risk, discontinuous updates, and essentially returning no information whatsoever.)

“[the model] requires information that can be only loosely estimated at the start of an epidemic, such as the proportion of infected people who die, and the basic reproduction number (…) rough estimates by epidemiologists who tried to piece together the virus’s basic properties from incomplete information in different countries during the pandemic’s early stages. Some parameters, meanwhile, must be entirely assumed.”

The report mentions that the team at Imperial College, which predictions impacted the UK Government decisions, also used an agent-based model, with more variability or stochasticity in individual actions, which require even more assumptions or much more refined, representative, and trustworthy data.

“Unfortunately, during a pandemic it is hard to get data — such as on infection rates — against which to judge a model’s projections.”

Unfortunately, the paper was written in the early days of the rise of cases in the UK, which means predictions were not much opposed to actual numbers of deaths and hospitalisations. The following quote shows how far off they can fall from reality:

“the British response, Ferguson said on 25 March, makes him “reasonably confident” that total deaths in the United Kingdom will be held below 20,000.”

since the total number as of April 29 is above 21,000 24,000 29,750 and showing no sign of quickly slowing down… A quite useful general public article, nonetheless.

terrible graph, again

graph of the day & AI4good versus AI4bad

Apart from the above graph from Nature, rendering in a most appalling and meaningless way the uncertainty about the number of active genes in the human genome, I read a couple of articles in this issue of Nature relating to the biases and dangers of societal algorithms. One of which sounded very close to the editorial in the New York Times on which Kristian Lum commented on this blog. With the attached snippet on what is fair and unfair (or not).

The second article was more surprising as it defended the use of algorithms for more democracy. Nothing less. Written by Wendy Tam Cho, professor of political sciences, law, statistics, and mathematics at UIUC, it argued that the software that she develops to construct electoral maps produces fair maps. Which sounds over-rosy imho, as aiming to account for all social, ethnic, income, &tc., groups, i.e., most of the axes that define a human, is meaningless, if only because the structure of these groups is not frozen in time. To state that “computers are impervious to the lure of power” is borderline ridiculous, as computers and algorithms are [so far] driven by humans. This is not to say that gerrymandering should not be fought by technological means, especially and obviously by open source algorithms, as existing proposals (discussed here) demonstrate, but to entertain the notion of a perfectly representative redistricting is not only illusory, but also far from democratic as it shies away from the one person one vote  at the basis of democracy. And the paper leaves us on the dark as to whom will decide on which group or which characteristic need be represented in the votes. Of course, this is the impression obtained by reading a one page editorial in Nature [in an overcrowded and sweltering commuter train] rather than the relevant literature. Nonetheless, I remain puzzled at why this editorial was ever published. (Speaking of democracy, the issue contains also warning reports about Hungary’s ultra-right government taking over the Hungarian Academy of Sciences.)

another terrible graph [about Neanderthalia]

Another terrible graph that misses by several orders of magnitude the terrible singularity of the United States in terms of homicides by firearms. And that does not include either the other [dove]tail, like the United Kingdom (0.7 per million) and Japan (0.1 per million).

bad graphics and poor statistics

Reading through The Guardian website, I came across this terrible graphic about US airlines 2016 comparison for killing pests pets they carry. Beyond the gross imprecision resulting from resorting to a (gross) dead dog scale to report integers, the impression of Hawaiian Airlines having a beef with pets is just misleading: there were three animal deaths on this company for that year. And nine on United Airlines (including the late giant rabbit). The law of small numbers in action! Computing a basic p-value (!) based on a Poisson approximation (the most pet friendly distribution) does not even exclude Hawaiian Airlines. Without even considering the possibility that, among the half-million plus pets travelling on US airlines in 2016, some would have died anyway but it happened during a flight. (As a comparison, there are “between 114 and 360 medical” in-flight [human] deaths per year. For it’s worth.) The scariest part of The Guardian article [beyond the reliance on terrible graphs!] is the call to end up pets travelling as cargo, meaning they would join their owner in the cabin. As if stag and hen [parties] were not enough of a travelling nuisance..!