Tuesday, October 17, 2006

Statistics 101: How to assess study validity

Case Example 1: UM's placement in world university rankings
"No no, the study is not valid. UM is better than UKM, and certainly on par with other world class institutions"

Case Example 2: 18.9% is the new 45%
"Our number is right, ASLI's number is wrong. Although I haven't looked at your study yet, I know your study methodology was flawed."

Lesson of the day- The study is deemed valid if:
1) The results demonstrated is the desirable and expected results, and
2) The deputy PM said so

The methodology used is totally irrelevant. The result becomes statistically significant if p less than 0.05 (p=pride). Best example on Dr. Liew's Blog. Tomorrow, we shall talk about studying the effects of zero gravity on teh tarik, roti canai and the inverse numerical relationship between spending RM 3.5 billion carrying out that study and our politicians' IQ.