Saturday, October 07, 2006

Caught Lying With Statistics

Mea Culpa. They suckered me, and I fell for it. I should know better, really, than to just take an Atlantic Monthly chart and post it just because it happens to support my fervent belief that America's schools need radical improvements in science and math education. A hat tip to Chad over at Uncertain Principles who tipped me off to the (now) obvious.

Some of you may remember the chart on relative test score performance of eighth grade science students from different countries (reposted here for your review.)

Looks pretty dismal, huh? Well, take a look at this re-plotted version of the test score chart from Chad's blog, which has been normalized on a scale of zero to one, as many physicists prefer. (Note that the original test score chart only ranged from 500 to 600.)


Okay, that doesn't look so bad. Now which should we believe? Is there evidence here for an international crisis in American science education?

This time, I actually did go all the way back to the original source, "Highlights From the Trends in International Mathematics and Science Study 2003." Surprisingly enough, despite all of the media teeth-gnashing, there was nothing in the report that actually defined how significant or relevant the differencesvbetween the test scores might be. The Atlantic chart accentuated the differences, and Chad's normalized chart de-emphasised the differences.

More investigation is in order. I'm going all the way back to the original reports to see what I might find out. Stay tuned...

1 comment:

Anonymous said...

I noticed that but did not want to offend... Phil