Friday, October 31, 2008

Six Sigma

The well-known management concept six sigma actually comes from the empirical rule that if one has six standard deviations (sigma) between the mean of a process and the nearest specification limit in a short-term study, there will be practically no items that fail to meet the specifications (i.e., 3.4 defective parts per million opportunities, corresponding to 4.5 sigma) in the long-term, assuming that the process is normally distributed.

Tuesday, October 21, 2008

Laplace Approximation

Laplace approximation is a simple and popular method that approximates a probability density function p(z) by fitting a Gaussian distribution to the local quadratic expansion of p(z) around its mode where p'(z) vanishes. One important usage of this technique is to compute the integral of p(z). A good explanation of this method from David MacKay can be found here.