ABSTRACT
Con rmatory factor analysis represents a different approach to using the common-factor model from anything presented previously in this text. Whereas earlier chapters concerned “exploratory” factor analysis, this chapter will brie y consider “con rmatory” factor analysis. Both forms of factor analysis are based on the same common-factor model. The difference between them is that whereas in exploratory factor analysis, the researcher seeks to discover, for a set of variables, common factors that account for their correlations, in con rmatory factor analysis the researcher begins with a hypothesis about what the common factors are and how they are related to the observed variables and seeks to test that hypothesis. In many respects, for the student, this may be a new way of doing statistics, because most forms of statistics taught to beginners in statistics stress estimation of parameters and then testing whether they simply differ from zero. A mean may be estimated and tested against the hypothesis that it is zero, or two means may be compared and their difference tested to see if it differs signi cantly from zero. Or one may compute a correlation coef cient to see if there is a relation between variables, and the estimated correlation is then tested to see if it differs signi cantly from zero. Only on occasion is the student shown that one can specify a hypothesis that, say, the mean is equal to 4.5, or the difference between means is 2.0, or that the correlation is .75 between two variables and test these hypotheses. In this second case, the stress is placed upon using substantive theory to provide values for these coef cients as hypotheses to test. In the rst, exploratory case, a null hypothesis of zero normally does not represent a speci c substantive hypothesis, since the researcher really believes there may be a nonzero value, but does not know enough to specify what it is, and if it represents a relationship, hopes that it is nonzero.
