I came across a classic example of "answering the wrong question" at a sugi conference.
A flue epidemic in Israel caused massive overcrowding of hospital wards, up to 200% at peak times.
A statistician from the health ministry built a regression model to predict "length of stay" based on demographic and previous health problems of patients.
The only problem was that she was measuring and attempting to forecast the wrong variable!
Was she wanted to measure was "how many people are in a hospital ward at any given time", though this variable is not independent of "length of stay" it is obviously not the same.
The hospilitization problem was not caused by people staying longer, but by a very large number of people, who could have had relativity short length of stays, but all these people got sick and needed to be treated at the same time.
The point I'm trying to stress, is that 80% of any project should be "defining what the problem actually is", so that the remaining 20% could be utilized in answering the right question.
From Tzippy
|