Most people have a pretty good idea of what extrapolating is – its where you look at a trend and predict what will happen if that trend persists. 

For example, if I said it took me 6 months to save £500, I can use extrapolation to predict how long it will take me to save £2000; its something we do all the time – yesterday I was driving down from Bristol, I could count off the the miles, and knowing the distance, I could predict if I would make it for dinner (I didn’t).

Scientists use this too. A good example is the way we can calculate the temperature of “absolute zero” by looking at the volume of a balloon as you heat it up. If you had a balloon at 25C, and you heat it to about 55C its volume would increase by about 10%. What does that tell us? It tells if we cooled it, it would eventually have no volume – and that this would happen at around -275C (-273.15C actually) – absolute zero.

Of course, the method relies upon assumptions – usually the assumption that the trend will continue in the same way (people often use the term “linear” to represent relationships that form straight lines when plotted on a graph).

What if the relationship is non-linear? For example, if little James is 5 feet tall when he is 10, how tall will he be when he is 20? Clearly he won’t be 10ft tall – that is because the relationship between height and age is “non-linear”.

Most of us are smart enough to extrapolate without knowing the jargon, but when the relationships get complicated a bit of maths and jargon can help.

For example, if we want to examine the population of bacteria in a petri dish, or the spread of a virus (or a rumour) through a population, our mental arithmetic is not always up to it. Luckily, some scientists have realised even these complex affairs have some predictability and although “non-linear”, they can still be modelled – graphs can be plotted and extrapolations made.

If this interests you, I refer you to books on epidemiology; I will move onto another sort of extrapolation – one used to check people’s theories by identifying ‘impossible’ extrapolations.

Let’s say, for example, that the want to predict  how the obesity epidemic will progress in the coming decades. If the media says obesity in a certain group increased from 14-24% between 1994 and 2004, and then goes on to predict that obesity will therefore reach 34% by 2014, does this withstand scrutiny?

Never mind that the definition of obesity may be faulty (BMI), never mind that they are extrapoliting from 2 data points – let’s rather ask if the linear trend is justifiable. This can be done by extrapolating the prediction to try to break it. 

If the model is right, obesity will go on increasing and soon enough 100% (or more!) of the population will be obese. This is clearly wrong – obesity is not likely to get everyone – vast swaths of the population are likely to be immunised to some extent against obesity due to active lifestyles and good dietary educations, or perhaps its in their genes, the lucky things. 

The truth will of course be more complex – the first group to become obese will be the most vulnerable, so an increase from 14-24% may incorporate that group, but each successively 10% will be harder fought.  All this is enough to suggest the predictions made for 2014 are doubtful, and those that go further are downright shameless. But it doesn’t stop them…

I am sure you can think of other suspicious trend-based predictions, like those for peak-oil or global warming. They could do with some improvements, so get to it!