# Accession Number:

## AD0671074

# Title:

## ASSOCIATION MEASURES OF BINARY VARIABLES: AN EMPIRICAL COMPARISON WITH RESPECT TO FACTOR STRUCTURES REPRODUCTION,

# Descriptive Note:

# Corporate Author:

## WASHINGTON UNIV SEATTLE DEPT OF PSYCHOLOGY

# Personal Author(s):

# Report Date:

## 1968-06-01

# Pagination or Media Count:

## 225.0

# Abstract:

The research begins with known factor structures on the basis of which data matrices are generated. These are converted to a number of binary matrices differing with respect to vectors of means. For each of these, several different types of matrices of association indices are calculated and a varimax-type factor analysis is applied. The criterion of effectiveness of an index of association is taken to be the adequacy with which it recovers the known factor structures. According to this criterion, the research indicates that 1 The matrix of association indices should be Grammian, since phi tends to be superior to other indices of association. 2 Sample size and distribution of underlying factor scores appear to have negligible influence on the relative effectiveness of any of the indices of association. 3 The phi coefficient is relatively less sensitive than other indices to disparity of means. 4 The tetrachoric coefficient, although in no case the poorest index of association, is seldom the best. Author

# Descriptors:

# Subject Categories:

- Psychology
- Statistics and Probability