No abstract

The measurement of agreement between ratings of patient symptomatology by two or more psychiatrists is discussed. It is noted that coefficients which allow for possible chance agreement are to be preferred, but they involve assumptions about the way in which chance factors may operate. Assumptions which involve prior probabilities of the incidence of a symptom appear to be too stringent, and a new coefficient, called the RE coefficient, is recommended in which is assumed to operate in a purely random way. Binary scales are discussed in detail, but methods of dealing with scales of wider range are also referred to.

The paper is concerned with the measurement of internal consistency of rating scales and interviewing schedules, with the assessment of bias between different raters and with coefficients for measuring the degree of agreement between them. Analysis of variance models are first employed, but reference is also made to earlier psychometric techniques and to recent work by Armitage et al. and by Fleiss.

SUMMARY.A factor analysis of the correlation matrix for the sub-tests of the WISC was carried out for each of the age groups 74, 104 and 134 years. In each analysis two factors were extracted and these, after rotation, are found to give patterns of loadings which are similar for all three age groups. The factors were then rotated : the first is labelled ' vg ' (verbal-intellectual), and the second ' sp ' (space-performance).Methods for obtaining factor scores are then given and it is argued that such scores are preferable to the cumstomary Verbal and Performance scores when summarising a child's performance on this battery of tests.