




1. We can easily compute this best linear fit for any data. Youcan compute the best quadratic or cubic fit also, by solving the given equations for k = 2 and k = 3.
2. It often happens that some data points are known with greater certainty than others, so that you want to make agreater effort to fit them than to fit others. This can be accomplished by giving different data points different weight,_{in the} sums defining the and . For example, you might give a weight, w_{i}, to the ith data point that is proportional to the reciprocal of the square of its expected error
_{ }
3. A problem with this method is that badly erroneous data points (called outliers) can greatly affect the answer. In practice you should look out for this. One way to do this is to check if the resulting curve changes _{ }significantly when the data points that are bad fits are removed from the data set.