Active Effects Selection which Considers Heredity Principle in Multi-Factor Experiment Data Analysis

The sparsity principle suggests that the number of effects that contribute significantly to the response variable of an experiment is small. It means that the researchers need an efficient selection procedure to identify those active effects. Most common procedures can be found in literature work by considering an effect as an individual entity so that selection process works on individual effect. Another principle we should consider in experimental data analysis is the heredity principle. This principle allows an interaction effect is included in the model only if the correspondence main effects are there in. This paper addresses the selection problem that takes into account the heredity principle as Yuan and Lin [23] did using least angle regression (LARS). Instead of selecting the effects individually, the proposed approach perform the selection process in groups. The advantage our proposed approach, using genetic algorithm, is on the opportunity to determine the number of desired effect, which the LARS approach cannot.