OPINION

Statistics, and why you can't always trust them

Statistics, and why you can't always trust them

Ninety-five per cent of people reading this did not know that today is World Statistics Day. OK, so I made that statistic up. As Mark Twain famously said: "There are three kinds of lies: lies, damned lies, and statistics."

Statistics are super important - they help us understand the world around us, allow for predictions and support decision making. In science, we perform statistical testing to help us interpret and present the data we collect.

Numbers don't lie. But unfortunately the way we collect, manipulate and present them can. So here is a very non-exhaustive list of just a few ways that statistics can be a little ... dodgy.

First, there can be bias in the way we collect data. Imagine I want to ask people how good they think my column is.

If I only ask my mum and my friends, most of them will say it's really good (or at least they better!).

This is selective bias - I've chosen a group of people who are likely to agree with me - instead of a random sample.

Then there's the questions you ask. If I ask you "what do you like about my column" I'm asking you a leading question - one that makes an assumption that you do indeed like my column, and biases the type of responses you'll give. A good survey will avoid leading or loaded questions.

Then there's the conclusions we draw from data. My favourite here is drawing false correlations. If you collect enough data on enough things, you will see some correlations between them, but it doesn't mean they are actually connected.

There is, for example, a correlation between the number of films Nicolas Cage has starred in and the number of people drowning in pools per year. Correlation yes, but causation? No.

We can also over-generalise from data. If I run a poll in my home town of Armidale, finding that 80 per cent of people think autumn is the best season, I can't conclude that 80 per cent of people everywhere think the same.

My results aren't going to be representative of the opinions of people in the rest of Australia, let alone the world.

I'm NOT saying we shouldn't trust statistics at all - after all, they're an important part of any research.

But this World Statistics Day, let's all try to be a bit statistics savvy. And remember that when the numbers seem a bit too good to be true, there's a good chance they are.

Dr Mary McMillan is a lecturer at the School of Science and Technology, University of New England