ARE OLIVIAS CAUSING VEGANISM?

HOW POORLY PRESENTED STATISTICS CAN IMPLY THINGS THAT AREN’T TRUE

MOST of us will have heard the phrase “correlation doesn’t equal causation” during our school science lessons.

That golden rule of statistics means that just because two things show similar patterns over time doesn’t necessarily mean that one causes the other.

However, when we see lines and dots on graphs that look similar it can be all too easy to fall into the trap of believing there is a meaningful link when there is in fact none.

A website called Spurious Correlations - set up by Harvard law student Tyler Vigen - details countless hilarious graphs demonstrating this fact.

For example, the number of deaths caused by drowning in a pool seems to match up with the number of films starring Nicholas Cage in a particular year.

Here we’ve created some of our own charts to highlight some common things to look out for when reading charts.

Firstly, it’s important to look for a statistical significance wherever possible.

None of these graphs encompass enough data points for the things they are saying to be statistically significant.

The link between girls called Olivia and Google searches for the word “vegan” only contains six pairs of data, for instance.

Usually statistical significance is denoted by something called the p value - if the p value is below 0.05 then we can say that there is some value in the pattern we’re seeing.

In the case of the Olivia graph, the p value was 0.3 - far higher than the value we need to say that something is true.

Secondly, it’s important to check the axes on the graphs.

Although it may look like there is a link between ecstasy deaths and the UK economy, there is actually no correlation at all.

That is because the axes and scales on the graph have been manipulated - one of them goes from minus six to four, while the other starts at zero and climbs to 60. That ought to ring alarm bells.

The lesson from these charts is to take care when reading data graphics.

You should ask yourself whether the graph has been designed to accurately reflect the data, or whether it has instead been designed to tell a story - perhaps a story that doesn’t quite stand up when you take a closer look.