Picturing the central limit theorem

The central limit theorem tells us that if we add data that contains random errors then we tend to get a sum that's normally distributed, independent of the distribution of the original data. I was wondering how many errors you need to add together to get something that looked close to a normal distribution, so I created some random data from a uniform distribution in Mathematica that represents such errors. I then and plotted one of these errors, the distribution of the sum of two of these errors and the distribution of the sum of three of these errors. Here's what I got.

Image001

One random error.

Image002

The distribution of the sum of two random errors.

Image003

The distribution of the sum of three random errors.

It certainly looks like you're fairly close to a normal distribution after adding together only three errors. I was surprised to see how quickly this happened.

Leave a Reply

Your email address will not be published. Required fields are marked *