## Tuesday, February 23, 2016

### A One-Track Mind

Today, as my stats students were working on a 2-sample t-test project determining if one version of a chapter test is statistically significantly harder than another, I planned a project for the next chapter.  My students have done the same project for the past few years now, one that involves the proportion of left-handed students at school, but I want to try something different this year.

I got a report of the birth date of every student in school and was able, without too much effort, to get just the birth month for each of our approximately 1600 students.  The project will be for students to determine if the proportion of students born in any specific month is about what we'd expect given the length of each month.  In other words, are births spread out evenly throughout the year?

Most of the months had numbers close to what we'd expect, but a bar chart of births per month showed September sticking out like a bump on a log.  I showed the chart and data to some nearby students, and almost immediately one boy said, "New Years."

Throw in Christmas parties and Christmas and I'm sure he was correct.

Wm Sears said...

Either that or he has fallen for the birth month fallacy. See

http://www.numberwatch.co.uk/extreme_value_fallacy.htm

Darren said...

Having read your link, this scenario does not appear to be a case of birth month fallacy. I haven't yet, though, run a "X-square goodness of fit" test on the data. When I get time....

Wm Sears said...

Here is an analysis of a larger number.

http://www.panix.com/~murphy/bday.html

Birth induction can also make holiday births less common.