Schools are awash with data, both quantitative and qualitative. Very little of this would meet the standards of a scientific journal, but we can apply some principles derived from scientific analysis to help us draw better inferences.
A lot of school data functions like a Rorschach inkblot. You can read all sorts of different things into it. This is similar in some ways to the scientific concept of p-hacking. If you have enough data then, by chance, you are likely to see something that looks like a pattern, but it is not a pattern because it arose by chance. The best approach is to draw such inferences tentatively and then, if you think they are worth exploring, use them to make predictions about future measurements. These can then be tested in a new collection of data.
For instance, if you trawl through your behaviour data and notice…
View original post 405 more words