Stats Explained

I often come across researchers that do something very silly, such as a study that correlates the distance from the patent office of the work address of inventors as listed on patents.

Why would they do this?

Well, firstly its cheap research because all that data already exists in the patent office databases which are readily available to the public. All you need is a student and a spreadsheet.

Secondly, they might be able to make some headline discovery that inventors are generally closer to the patent office than expected.

Apart from publishing this in the Japanese Journal of Inexplicable Analytics, this type of stuff is gobbled up by the editor-free internet news and info sites.

Thirdly, they may make calls (more newsworthy efforts) for patent offices to be distributed so that all inventors have equal rights and aren’t being geo-positionally discriminated against.

Fourthly, they will get to talk at international conferences on IP analytics where they will confound everyone with their hitherto unexpected results.

And, finally, based on all that they might get grants to do it all again, and again, and again.

Imagine a perfect storm; a nanny state with politically correct agendas where policy is developed by sheltered-workshop public servants that couldn’t spell either correlation or causation.

They’d be straight onto it; developing plans to distribute the patent office from its mausoleum in Canberra. They’d probably put one person in each of the residual post offices in the land.

Which isn’t a bad idea really; the post offices already provide about 200 odd government and commercial services. What harm could one more do? Just don’t try to buy a stamp.

Of course, with any understanding of statistics your researcher could have tested whether the correlation had any causation in any number of ways.

For example, by digging further into the data to find the patenting habits of inventors whose companies had moved offices at one time or another.

Of course the researcher would find that the rate of inventorship wasn’t at all changed by the change of domicile, and therefore he or she could happily conclude that the original correlation was just one of the likely co-location of both businesses and patent offices in the limited number of commercial real estate locations.

The odd, odd, odd thing is this; perfectly well trained researchers, when confronted with the headline correlation, just wouldn’t question it, or they wouldn’t know how to.

If I was put in charge, all Donald-like, one thing I would do, I can assure you, is set up some sort of torture gulag for researchers-in-training that fluffed the sanity clause, as per above.