So, let me assume that my best fit linear function has the form: Let me generically label the data as ( xi, yi ). So, I can write di and di2 as: Well that is just great. Now what?