a note on bivariate people
When ArtO gave a talk at MMDS, he showed a slide that featured a seemingly-odd distribution of genome-influenced data.
Why? he asked, were there two clusters distributed bivariate-ly, with outliers in the corner of the quadrant he made visible?
I didn't tell him at the time, but my guess was that it has to do with the bivariate distribution of people, and in perhaps nature, in general.
Take thing-orientation and people-orientation. Spacetime and Facetime if you like. Simon Baron-Cohen (oh how I love to transpose your name) notes, kind of, that Spacetime and Facetime computational complexity are two ways of categorizing people.
His research has been, well, understood very differently from the way he imagines it, I think, as people tend to think about Spacetime and Facetime as points along a single dimension, a single continuum.
SB-C notes that it helps if you think of Space and Face (thing and people) as two different dimensions. Fine and dandy, but people still don't grok the geometric nature of probability theory, and they end up implicitly projecting along a single dimension (the line perpendicular to y=x which intersects the whatever-median of our people probability-distribution), as most people are somewhat categorizable along that simple bisection.
Thing is, there's other people two! Err, too.
Think instead of the probability space as hilly. You have the spacetime people on one big hill up in the Northwest, and the facetime people on another big hill down somewhere in the Southeast. Now, I don't know what to call the people who know neither space nor face, but there are lots of people left out in between the two big hills.
Not only that, but there's yet another quadrant! A small community of spacefacetime-grokking folks up in the Northeast. There's not so many of them, sure, as it's unlikely to be blessed, so it seems, with the ability to compute in spacetime /and/ facetime -- spacefacetime that is -- and thus these folks, well, often get screwed, as they don't know where to go and don't know where they came from... Neither fish nor fowl, alpha nor omega, neither unsaturated nor saturated.
I think, though, it's time to take back the bivariate in bi-* of all stars and stripes, and to reject,partially randomly at least, once more, this nasty bass mass of the deterministic PSPACE hierarchy.
Why? he asked, were there two clusters distributed bivariate-ly, with outliers in the corner of the quadrant he made visible?
I didn't tell him at the time, but my guess was that it has to do with the bivariate distribution of people, and in perhaps nature, in general.
Take thing-orientation and people-orientation. Spacetime and Facetime if you like. Simon Baron-Cohen (oh how I love to transpose your name) notes, kind of, that Spacetime and Facetime computational complexity are two ways of categorizing people.
His research has been, well, understood very differently from the way he imagines it, I think, as people tend to think about Spacetime and Facetime as points along a single dimension, a single continuum.
SB-C notes that it helps if you think of Space and Face (thing and people) as two different dimensions. Fine and dandy, but people still don't grok the geometric nature of probability theory, and they end up implicitly projecting along a single dimension (the line perpendicular to y=x which intersects the whatever-median of our people probability-distribution), as most people are somewhat categorizable along that simple bisection.
Thing is, there's other people two! Err, too.
Think instead of the probability space as hilly. You have the spacetime people on one big hill up in the Northwest, and the facetime people on another big hill down somewhere in the Southeast. Now, I don't know what to call the people who know neither space nor face, but there are lots of people left out in between the two big hills.
Not only that, but there's yet another quadrant! A small community of spacefacetime-grokking folks up in the Northeast. There's not so many of them, sure, as it's unlikely to be blessed, so it seems, with the ability to compute in spacetime /and/ facetime -- spacefacetime that is -- and thus these folks, well, often get screwed, as they don't know where to go and don't know where they came from... Neither fish nor fowl, alpha nor omega, neither unsaturated nor saturated.
I think, though, it's time to take back the bivariate in bi-* of all stars and stripes, and to reject,
posted by L. Wu @ 9:49 AM
1 Comments:
At 11:52 AM,
L. Wu said…
Perhaps this is more precisely explained as a discretized Pareto-like bivariate distribution (!).
This explains the flawed reasoning behind false dichotomization and accounts for statistically improbable NE quadrant outliers, while also giving weight to a large centrist population.
Post a Comment
<< Home