I have been looking at the paper `A Critique of Software Defect Prediction Models`. This is a paper mentioned in the reading list. It has been a good read. Since I was involved in the CMM effort at the Bank of America I can see many of the mistakes made then by reading this paper. Metrics about defect prediction are difficult. The unknown relationship between defects and failures makes this problem a pretty large one. The mistakes often made where ones I have seen being done at the Bank of America. The ones mentioned in the paper have all been done their. --the problems with `multivariate statistical approach, problems of using size and complexity metrics as sole predictors of defects, problems in statistical methodology and data quality, and false claims about software decomposition and the goldilock`s conjectures, are all problems exhibited in the Bank of America`s software process defect predictions. `Management by Fact` has a loose definition of the word `fact` in their process. At any rate I understand because this stuff is hard. One of the questions on the test is related to self experience with these problems. I think that I have the ammunition. I will just need to be ready to cover the right bullet points.