Fig. 1. Classes and their inheritance relations for the 5-level version of the “Boerse” program.

Table 1. Overview of our experiments (N -G, N -U) and the previous experiments (P -1a, P -1b/-1r, P -2, Cartwright replication C)

Table 2. Characterization, by program version, of the typical solution effort for the tasks (the task used in P -1b and -1r and the Cartwright replication are similar to that of P -1a)

Investigated methods: number of methods that must be analyzed and understood for solving the problem. Hierarchy changes: number of times one must switch to a subclass or superclass during that method understanding process if it is performed by dynamic tracing (i.e., following the execution call sequence). Solution methods: number of methods that are copied from existing classes (verbatim or with changes) or modified in order to create the solution. No methods needed to be written from scratch.

Table 3. Various prediction models for average code maintenance times

t is the time in minutes for the “add class” task, h is the number of hierarchy changes during understanding, m is number of methods to be understood, exp=H and exp=L are the groups with high or low levels of experience, d is the inheritance depth (ID), k is the number of coefficients (each model has 16 minus k degrees of freedom), r² is the fraction of variance explained by the model, p_max is the p-value of the least-significant coefficient in the model. Except for all coefficients d, all coefficients are significant in all models. 14[exp=L] denotes a separate constant offset of 14 for the low-experience groups only and 8.9m_exp=L denotes a coefficient for m that is 8.9 for the low-experience groups and 0 otherwise (an interaction term).