A Comparative Research on Data Analysis with Factorial ANOVA, Logistic Regression and CHAID Classification Tree Methods

Abstract

When the data structure is large and complex, the extraction of information hidden within the data is called data mining. In the context of data mining, there are numerous methods developed for statistical data analysis. When these methods are classified as conventional-classical methods and current methods, factorial ANOVA (FANOVA) and Logistic Regression (LR) methods are shown as conventional methods, while decision trees called Classification Tree (CT) and Regression Tree (RT) can be shown as current methods. The method to be used in statistical data analysis is directly related to the researcher’s hypothesis (i.e. purpose) and variable type. Therefore, the choice of data analysis method is important. In this regard, studies in which methods are examined comparatively are guiding. In this study, a dataset on which inferences could be made by ANOVA, LR, and CT methods was analyzed.With this dataset, the relationship between the birth type (single-twin) (dependent variable) and the yield year and maternal age (independent variables) in an Awassi sheep flock was examined. The findings of each method were interpreted in its own specific way. The methods were compared in terms of model fit, the similarity and differences of the information that they presented, and explaining the relationship between the dependent and the independent variables. It was concluded that each method offered different inferences based on purpose and perspective. It is believed that it is the right approach for researchers to determine the data analysis method appropriate to their goals by taking into account the data structure.