Predicting Numbers of Problems in Development of Software
- Created: Tuesday, 01 November 2005
1. Take raw data in the form of problem reports (PRs), including the dates on which they are generated.
2. Remove, from the data collection, PRs that are subsequently withdrawn or to which no response is required.
3. Count the numbers of PRs created in 1-week periods and the running total number of PRs each week.
4. Perform the interpolation by making a least-squares fit of the Weibull function to (a) the cumulative distribution of PRs gathered thus far and (b) the goal of no more PRs after the currently anticipated launch date. The interpolation and the anticipated launch date are subject to iterative re-estimation.
This work was done by Charles H. Simonds of Lockheed Martin Corp. for Johnson Space Center. For further information, contact the Johnson Innovative Partnerships Office at (281) 483-3809. MSC-23532