| This work encompasses three primary categories: (1) development and evaluation of an information theoretic measure modeling control-flow nesting and control-flow expressions, (2) statistical analyses of several software measures (some traditional and some from our model) applied to sets of functions making up fourteen large, professionally written, public domain software systems written in the C programming language, and (3) statistical analyses of the relationships of the measures to the number of changes per function on successive versions of the software systems.; We develop a stack-based Markov source model that uses the stack top in conjunction with the current token to model the conceptual complexity of C programs. This model assumes that a program is a stream of program elements. We evaluate the model on sets of functions making up fourteen large, professionally written, public domain software systems.; We also perform statistical analyses of various traditional software measures, transformed traditional software measures, and software measures based upon our information theoretic model generated on the sets of functions. Using statistical analysis techniques, we document answers to exploratory questions similar to: (1) ‘What can we learn about the data (e.g., distributions, characteristics, etc.)?’ (2) ‘What is the typical behavior?’ (3) ‘What are the descriptive (i.e., summary) statistics?’ (4) ‘What are the relationships?’ and (5) ‘What graphical aids best display the behavior of the data?’.; Finally, we investigate the relationships of the various (traditional, transformed traditional, and model based) measures to the number of changes per function on successive versions of the software systems. We evaluate the measures as indicators of change prone functions. |
