pull down to refresh
the log-log[1] "Posts vs Sats Stacked" [2nd row, left side] shows it beautifully
linked in case people like wikipedia; tl;dr, switching both scales of a graph from linear to logarithmic causes some kinds of data behavior to "jump out" because the log scales turn certain kinds of curves into lines of recognizeable angles ↩
I'm being facetious, cuz all the power-law log-log stuff is complete bullshit (#1433264).
Nice to find a Pareto distr in the SN wilderness
honestly I'm not well-versed in the detailed differences.
the line on the log-log chart, with meta outlying, is undeniable.
I'm being facetious, cuz all the power-law log-log stuff is complete[1] bullshit (#1433264).
I agree that quibbling over whether some fitted curve is n^x, x^n, or something else, is futile in most situations... although a conversation in one situation could be complete bullshit, without reducing the abstract accuracy of the mathematics.
emphasis mine, on the word that pissed me off ... ↩
I thought Pareto distributions show up everywhere
unless you[2] metrize[3] by something absolute, then the distribution of distributions remains hopelessly subjective[1]
i.e., whichever question interests anyone, will reveal more about where specifically data were collected, and how, rather than about isotropic properties of the data-space ↩
"you" could be part of the data-space, although obviously it's easier to remove bias if the study designer isn't navel-gazing ↩
the "default metric" in this case would just be radial density, looking out from one subjective point, into the surrounding data-space; ideally, averaging that across multiple points; theoretically, one could hope to move beyond this limited frequentist cloud, although I haven't encountered anything to suggest humanity has done this yet. ↩
Yahyah, a little bit
Oh look, it's a power law!!