[LINK] "The use and abuse of statistics"

[rfmcdonald]

Razib Khan's article at The Guardian's Comment Is Free is worth reading in its entirety.

It is an old trope that one can lie with statistics, but it might be more accurate to admit that selective use of numbers can easily mislead. Introductory biology students often learn that one bacterium in a petri dish will clone itself so that if it kept going at the same rate it would in short order fill up the universe. But common sense soon tells them this won't happen. Projecting from the first generations is folly.

Human societies are infinitely more complex in their structure than bacterial colonies, so one should be exceedingly cautious about games of prediction. Across the west the 1960s was a period of cultural change as church affiliation dropped precipitously. If one extends that decline out from the interval 1960-1970 Christianity should be extinct, and yet it is not. The decline in religiosity slowed by around 1980. Why? The will of God? A more plausible explanation is that social pressures which enforced religiosity before 1960 no longer operate, so those who were never particularly devout are more honest with themselves and society.

[. . .]

The universe of statistics is vast, so polemicists can choose congenial data to "prove" their trends. Labels often hide more information than they reveal. Prophecies of the extinction of religion, or its total ascendancy, inevitability fall prey to the weaknesses of linear extrapolation. The most important thing that science can tell us about most trends is that they will some day reverse.