Ανάλυση πολυδιάστατων πινάκων συνάφειας με διατάξιμες μεταβλητές ταξινόμησης
Παπαδοπούλου, Μαρία Α.
The development of methods of data analysis had started after the end of the second world war. The beginning of the study of (ordinal) categorical data and especially an attempt of modeling them, parametrically and non parametrically, is placed at sixties. Leo Goodman is considered as a main contributor toward this direction. Serious effort has been made in the field of graphical representations of the structure and the relationships underlying categorical data. In this thesis, we focus our interest on the analysis of (ordinal) categorical data for multiple-way contingency tables. In the first chapter, a history of analysis of categorical data is represented and in the second chapter, we build graphical representations for categorical data, i.e. mosaics and association plots. In the third chapter, three variables loglinear association models are presented and also, an algorithm for model selection. In the forth chapter, we define association graphs and a brief report for graphical models is given. In the fifth chapter, we discuss latent variables models and in the sixth chapter, we present homogeneity analysis (or multiple correspondence analysis) and a detailed example.