Nick logo Credibly Curious

Nick Tierney's (mostly) rstats blog

2013-09-01

Statistics Can Do Anything?

Categories: Statistics Blag

2 minute read

Sometimes people are fond of saying:

Well, you know, the thing with statistics is that you can get the statistics to say whatever you want.

I think that this is the equivalent of saying:

…What I am saying is true because of my argument.

You can say that anything is true, but what really matters is the strength of your argument.

And statistics…

The same goes for statistics, you could perform a variety of tests to support a particular point.

However, your argument depends on the suitability of your statistical test for what you are trying to prove or demonstrate

Statistics are used to articulate a particular point. The author has the responsibility of ensuring they are correctly using a statistic to prove this point.

The reader also has a responsibility to evaluate their use of tools to construct and support their argument.

So yes, statistics could be used to support the argument

Johnny steals tim tams from Nick, as he has done this a statistically significant number of times.

Upon reading this, one should be very keen to investigate the ways in which the statistics were used, and ask questions like:

What data was it? Observation? Who observed it? How did they observe it?

What sort of anaysis was used? Why?

Were any data transformations used? Why?

And so on.

So when someone says that statistics can be used to prove anything, I like to replace this with:

You can argue any point, and sometimes the people who don’t know any better will nod their head quietly and whisper statistics said it was so

Others will nod their heads too.

But it is your job as a good researcher / reader to ask the all important:

Why?