Friday, November 06, 2009

When to close a school: a Japanese study that might work here


Some time ago, I saw a link to a study that used hard data to predict when an epidemic is likely and so when a school should be closed.

I should point out that I don't know the exact definition of 'epidemic' - the study uses data to predict if an absentee rate of 10% of students was likely. Losing 10% of students sounds like an epidemic to me, so that's how I am using the word.

The researchers looked at historic absentee rates in previous flu seasons. When they found the magic number of 10%, they looked at the week previous. a single day of 5% absenteeism was a good predictor of 10% absences a week later, but two consecutive days of 4% was better and three days of 3% was better still.

This graph is from a review of the article. I don't know why the numbers are different.
Algorithm.jpg

In the study, absences were known to be due to the flu. This might be a problem at my university as students often use a simple prescription form as evidence of illness as an excuse for missing class. They should be using the official excuse letter (Korean "Hal-ae-sa" or "Seo-yo-sa") but many teachers are lenient. This means that some students might be faking the illness but also that the university might not be aware the student was sick.

I'm not satisfied with this post - my son wants some more attention, so I must turn away from the computer. Oh, not 'must turn away', 'get to play with my son'. I may look it over again later, but no promises.

No comments: