Further Causal Search Analyses With UCC's Effort Estimation Data

Abstract

Correlation does not imply causation. Though this is a well-known fact, most analyses depend on correlation as proof of relationships that are often treated as causal. Causal search, also referred to as causal discovery, involves the application of statistical methods to identify causal relationships using conditional independences (and/or other statistical relationships) within data. Though software cost estimation models use both domain knowledge and statistics, to date, there has yet to be a published report describing the evaluation of a software dataset using causal search. In a previous paper, the authors ran a PC causal search algorithm on Unified Code Count's (UCC's)1 dataset of maintenance tasks and compared them to correlation test results. This paper builds on the previous paper to introduce causal discovery to software engineering research by exploring additional causal search algorithms (PC-Stable, fast greedy equivalent search [FGES], and fast adjacency skewness [FASK]) and comparing their results to the traditional multi-step regression analysis.