February 16th, 2009

get your own box

facs and figgers

i am reading stumbling on happiness, by daniel gilbert. it's interesting, but i find myself even more suspicious of the descriptions of psych studies than i usually am. often i'll read or be told about something, like "they did x and the test subjects responded with y, therefore z" and my first reaction is "bullshit! that could mean this that or the other, depending on all these other unmentioned factors". now, likely (i hope), the actual studies took that into account -- i don't feel inclined to go wading through academic articles to find out -- but the recountings of them are just not convincing in the least. maybe this means i should suck it up and read the actual studies if i want academic rigor, or i should suck it up and live with ambiguity if i don't bother.

however! the real reason i am here is a shameless misuse of statistics that i just found, so off it prompted me to come post this. gilbert sez "the average american... marries more than once". now according to the endnotes (because i'm a nerd who checks endnotes), the actual numbers are 10% will never marry, 60% will marry once, and 30% will marry more than once. now, yes, the average american according to this will marry at least 1.2 times. but that is nonsense, you can't marry .2 times: you is or you ain't. the number needs to be rounded. if 99.9% of people married exactly once and 0.1% married more than once, on average we would all marry more than once. but saying it that way gets the situation exactly backwards. whoopty doo.

it's funny that, in a book dedicated in part to debunking the ways people unconsciously cook facts to support their beliefs, he'd pull such a shenanigan.
  • Current Music
    Nine Inch Nails - And All That Could Have Been - 6 - The Day the World Went Away