多水平模型(英文原著) chap2.doc

Chapter 2 The basic linear multilevel model and its estimation 2.1 The 2-level model and basic notation In this chapter we introduce the 2-level model together with the basic notation which we shall use throughout the book. We look at alternative ways of setting up and motivating the model and introduce procedures for estimating parameters, forming and testing functions of the parameters and constructing confidence intervals. To make matters concrete, consider the following data. It is a dataset we shall use again and it consists of 728 pupils in 50 primary (elementary) schools in inner London, part of the Junior School Project (JSP). We consider two measurement occasions: the first when the pupils were in their fourth year of schooling, that is the year they attained their eighth birthday, and three years later in their final year of primary school. Our data are in fact a subsample from a more extensive dataset which is described in detail in Mortimore et al (1988). We use the scores from mathematics tests administered on these two occasions together with information collected on the social background of the pupils and their gender. In this chapter the data are used primarily to illustrate the development of basic 2-level modelling. In chapter 3 we shall be studying more elaborate models which will enable us to handle these data more efficiently. Fig. 2.0 is a scatterplot of the 11-year-old mathematics test score by the eight-year-old test score. In this plot no distinction is made between the schools to which the pupils belong. Notice that there is a general trend, with increasing 8-year scores associated with increasing 11-year scores. Notice also the narrowing of the between pupil variation in the 11-year score with increasing 8-year score; an issue to which we shall return. In Fig. 2.2 the scores for two particularly different schools have been selected, represented by different symbols. Two things are apparent immediately. The school represented by th