Di Nuovo, A. G., Di Nuovo, S., & Buono, S.
Missing data analysis: A comparison between statistical methods and a fuzzy clustering algorithm
In psychological studies, and more generally in scientific research, it is frequent to have data lacking some variables in the sample to analyze. This lack, if not faced in a correct way, could involve a weakening of the research validity, especially if the sample is small. In this article we introduce the solutions provided by the most popular statistical softwares and one new technique based on fuzzy logic. In order to exemplify the methodology and to highlight the specific characteristics of every solution, a comparison between these methods was made on a data-base of 186 adult participants mentally retarded. Results demonstrated that completion techniques, and in particular the one based on fuzzy clustering, could lead to an effective approximation avoiding at the same time sample size reduction, which decreases the power of the effects found.
