There’s a great article here on estimating correlations and means when you have missing data in your datasets. It uses the Expectation Maximisation algorithm to calculate the missing values, and what is interesting is how much the implied correlation changes after applying EM.
Garth Sundem of Geek Logik fame has a blog in which he comes up with formulas for wrapping every-day decisions in a mathematical framework. It’s something I’ve been thinking about for quite a while – especially with regards to procrastination. The trick is to quantify all the factors that you think are important in the issue you are considering, and then describe the relationship between the variables mathematically.