Agreed ... And you're right - I got more than a chuckle out of reviewing the data presentations.
Somewhere there are pages and pages of those graphs!
When I was doing research on the ONS data, for practice we'd choose the daftest variables we could find to combine in a graph, like, I dunno, dog ownership+magazine subscriptions+fear of street crime to find out whether those aspects of people's lives correlated, and then we'd come up with ideas about why they did or not.
So... lets say all 3 correlated nicely. Why?
People who owned a dog might be more outgoing than non dog-owners because they need to walk the dog and would naturally get to know other dog-owners. They might feel less afraid when they go out because they'd expect the dog to guard them.
Subscribers to magazines might be more literate and educated than non-subscribers so might be more affluent and able to afford a car rather than walking everywhere when they aren't with the dog.
Would this make them less afraid of street crime? At first glance if all three variables corresponded - they made similar lines on the graph - it would seem that they did. You'd have to go back and ask a lot more questions to be sure.
It was a statistical exercise rather an attempt to find a real meaning; and we did move onto more serious efforts, but it taught us a lot about the old
spurious correlations.
