Effort Estimation Using Scott Knott Test

Software Cost Estimation is used for large-scaled and complex software systems leads managers to settle SCE as one of the most vital activities that is closely related to predicate the success or failure of the whole development process. Propose a statistical framework based on a multiple comparisons algorithm in order to rank several cost estimation models, identifying those which have significant differences in accuracy, and clustering them in non-overlapping groups. In the existing work Scott-Knott test was used to rank and cluster the software estimation models. The test proposed by Scott Knott, a procedure of means grouping, is an effective alternative to perform procedures of multiple comparisons without ambiguity. This study aimed to propose a modification related to the partitioning and means grouping in the said procedure, to obtain results without ambiguity among treatments, organized in more homogeneous groups. In the proposed methodology, treatments that did not participate in the initial group are joined for a new analysis, which allows for a better group distribution. The proposed methodology is considered effective, aiming at the identification of elite cultivar groups for recommendation.