Modelli log-linearigraflici-catena per l'analisi della durata di intervalli tra nascite

Abstract

Log-linear graphical chain models have a one-to-one correspondence to directed graphs where any missing eUge represents a conditional independence statement. Hence, they can be considered in the study of directional links among categorical variables. The meaning of these models, following an asymmetrical approach in the analysis of demographical data, is presented. Making use of WFS (Italy, 1979) data referring to the lenght of two birth intervals (first and between first and second birth), some applications of log-linear graphical chain models are performed. The structure of the relationships among background variables, intermediate variables and birth interval lengths is described.