Weighted least squares

Weighted least squares is a way of making predictions that takes into account certain information so that certain measurements can be given more importance than others. Imagine if you had to guess how much candy somebody had in their bag. You could guess a number, but what if you found out they had mostly small candies? Then you should probably guess a smaller number because they can't fit as many large candies into the same space.

Weighted least squares works in the same way; it pays more attention to measurements that are more reliable, and less attention to measurements that are less reliable. In this example, the quantity of small candies is more reliable than the quantity of large candies, so it gives more weight to the small candies and less weight to the large candies when making a prediction.