What statistical test to compare data sets?

lynntlau · Post by **lynntlau** » Thu Jan 18, 2007 6:54 pm

Help! I'm teaching Grade 7 science and I've never taken a statistics course!

I'm wondering what statistical analysis my students should use when comparing two data sets. Their science fair question is: Do Grade 7 boys have a faster reaction time than Grade 7 girls? My students are 13 years old.

Thanks!

LYnn

deleted-71447 · Post by **deleted-71447** » Thu Jan 18, 2007 7:48 pm

You could use Student's t-tests:
http://www.physics.csbsju.edu/stats/t-test.html

MelissaB · Post by **MelissaB** » Thu Jan 18, 2007 7:53 pm

The proper statistical test depends on the distribution of the data. If it's distributed normally (both sets of scores look like a bell-shaped curve) an unpaired two-tailed t-test is the right test to do. (Technically, your hypothesis is one-tailed, but the convention in science is to always do two-tailed tests because they're more conservative.)

If it's not, off the top of my head I think the nonparametric equivalent is a Wilcoxon test, but I'd have to look it up.

Chances are your data will be distributed normally enough (even if they aren't perfect bell-shapes) that a t-test will be appropriate--and since it's only 7th grade you don't really need to get into the difference between parametric statistics and nonparametric statistics.

As for how to do one--what statistical program do you have? I /think/ if I remember correctly you can actually persuade Excel to do a t-test if you don't have any stats programs, but I'd have to remember how to do it. It's easy to do in most stats packages. There's a free one called 'R' on the web that's supposed to be good, but from what I hear it's not especially user friendly.