sjk

• Oct 20th, 2007

 

Metrics analysis is done once the trend is plotted. There are certain rules. 7 points on one side of mean line, 5 points with an ascending / descnding trend, points outside LCL UCL are few of them. When there are any deviations, causal analysis needs to be done. Fish bone diagram would be useful for this.

dilton_dalton

 Profile Answers by dilton_dalton 

• Dec 9th, 2007

 

What SJK described are some aspects of statistical process control (SPC) or statistical quality control (SQC).  While SQC depends on the analysis of data (normally defect data), these are not the only data that need to be analyzed. 
Analysis means to take something apart to examine its components.  Thus chemical analysis determines what elements and chemicals something is made of.  The word means the same when applied to metrics.
Metrics are measurements.  A software development project is an experiment that is based on a number of assumtions, theories and estimates.  To confirm these, measurements are taken for all significant processes and then analyzed.  The results of the analysis are used to make decisons.  For example, if the assumed code productivity for a project was 800 lines per person month and the actual measured average is 300, it is probably not a good idea to use 800 to estimate the next project.  Determining this average is a metrics analysis task.  The effort data might show that a lot of effort was spent fixing problems during system test.  The data might indicate that this could be reduced by spending more effort on code inspection.  Confirming this relationship is a metrics analysis task.