Jay Mathews reviews Gerald Bracey's new book, Reading Educational Research: How to Avoid Getting Statistically Snookered:
As a popular writer and speaker, with regular columns in two monthly education magazines, the Phi Delta Kappan and Principal Leadership, and acidic annual reports on the condition of public education, Bracey has been exposing statistics abuse for years. But I have never seen him put together all that he knows as well as he has in this book...Here is a good example of the Bracey passion for clarity. He is addressing the difficult concept of correlation, a key to many misunderstandings of educational statistics and to most bad education stories, including some written by me:
"We can correlate any two variables. Whether or not the resulting correlation makes sense is another question. Before everyone started wearing jeans, the Dow Jones stock market index correlated with skirt length. Shorter skirts were associated with good economic times and a rising market. Longer skirts were correlated with recessions. To the best of my knowledge, no one suggested raising hemlines as a means to boost the stock market. Similarly, there is a correlation between arm length and shirtsleeve length. Given ONLY a correlation coefficient, though, it makes as much sense to think that increasing sleeve length will make arms grow longer as it does to think that longer arms will mean longer sleeves. In this case other information could be adduced to assist in determining which way the causal relationship would operate"...
Bracey is a prolific and aggressive critic of No Child Left Behind and the rising use of standardized tests to assess schools and students, but he is too careful an analyst to embrace the most popular alternatives to testing without also giving them the third degree. One favorite of the anti-testing movement, portfolios (samples of student work), is seen by Bracey has just another idea with problems. So you have a nice big portfolio envelope, Bracey says. What do you put in it? "Typical work or the best work?" Bracey asks. "Who decides what is best? Teacher or student?" What do you do, he asks, when teachers disagree about the quality of the work?
All excellent points. Sounds like a good read to me.
His best suggestion is adding courses in what he calls "consumer-oriented probability and statistics" to our curriculums.
Hey, he stole my idea! I always thought a statistics course would go over quite well in high school, if it were retitled to be, "Bulls**t Detection 101."
Posted by kswygert at December 21, 2005 10:43 AM