Stats Made Easy

Practical Tools for Effective Experimentation

Saturday, February 21, 2009

Statistics on education and education on statistics

“Armed with a bevy of State testing data” our local school district’s Superintendent dispelled the notion that Minnesota charter schools provide a superior education. Charter schools are publicly funded, but they are run by independent boards. Their enrollment has doubled in the last 5 years, thus inciting battles of statistics like the one presented by the Superintendent this week. His main point was that published assessments do not account for the impact of special education, "at-risk" and other students that do not attend the particular charter school in our area. (To be fair to the ideal of charter schools, some actually provide for students with particular needs that would otherwise get lost in the shuffle.)

In the interest of helping the general public sort out the endless conflict of statistics by opposing groups like this, the Union College Academy for Lifelong Learning (UCALL) in Schenectady, New York invited an illustrious faculty, including Gerald Hahn – formerly statistician emeritus from General Electric, to teach a course “Numbers in Life.” This 10 hour presentation was really all about statistical literacy, but ironically the course coordinator advised that the word “statistics” in the course title would be a turn-off. That tells it all: How can people be educated a on a subject that must not be named (like the evildoer in Harry Potter)?

To the everlasting credit of Hahn and his fellow teachers, they gave “Numbers in Everyday Life” the old college try. Here are some of the course take-aways published in the February AmStat News (the parenthetical comments are my own):
-- Always ask who is taking/reporting the numbers and how they obtained.
-- Be wary of advocate’s numbers (such as a glowing report on a drug study sponsored by the manufacturer).
-- Remember that the news media seek surprising numbers.
-- Appreciate limitations of observational studies and differentiate correlation from causation.
-- Controlled experimentation (the forte of my company Stat-Ease) is the gold standard.

By the way, you’d best be wary of how I cherry-picked these take-aways to support the cause for design of experiments. Actually, this was one of the take-aways that I conveniently omitted (“cherry-picking”). ;) See the entire AmStat story posted via the YUDU epublishing initiative.

“One of the greatest contributions of statistical thinking… is design of experiments.” - Gerald Hahn


Post a Comment

<< Home