Factor Invariance in the Measurement of Children's Manifest Anxiety

Abstract

Anxiety has been shown to be an important variable affecting an individual's performance ( 4 ) . In measuring anxiety, an important problem is the stability of the factors across populations. That true factor scores for individuals remain constant independent of the selection of tests or the sampling of individuals in a set of experiments is the definition of factor invariance. The purpose of this study was to determine the factor invariance of the Children's Manifest Anxiety Scale, which ( 1 ) has been adapted from the scale for adults ( 6 ) . It has 42 statements measuring manifest anxiety and 11 items for an L scale (tendency to falsify). The score on the Anxiety factor is derived by counting the number of Yes responses. The Lie score is determined by counting Yes responses on 9 items and No responses on 2 items. Kitano ( 3 ) reported Kuder-Richardson reliability of .86 for the Anxiety Scale and found significant differences between regular classes and special ones on adjustment as evidence for discriminant validity. For the present study, Ss were 611 pupils in Grades 4 through 7 of 2 elementary schools of a southeastern suburban school district. They participated in research to determine the effects on manifest anxiety of moving into a vastly different school environment with few interior walls, no windows, and much more open space ( 5 ) . The 53 individual dichotomous-item responses of the 611 pupils were factor analyzed by use of the BMD03M program ( 2 ) . Principal component solution and orthogonal rotation of the factor matrix were performed with three iteration cycles. Two factors were identified? Factor consisted of the 42 items identified by the authors as the Anxiety Scale; this factor had an eigenvalue of 14.84 and accounted for 28.0% of the variance. Rotated factor loadings for individual Anxiety items ranged from .44 to .69 with a median of .59 and a mean of .59 by Fisher z-transformation. (Two items had the highest loadings: I worry about what is going to happen; I feel that others do not like the way do things.) Factor I1 consisted of the 1 ]terns originally identified as the L scale; the eigenvalue was 3.51 and accounted for 6 6% of the variance. Factor loadings ranged from .34 to .82, with a median of .65 and a mean of .69 by Fisher z-transformation. The highest loading was on the item: I am always nice to everyone. The factor structure for chis sample corresponded precisely to that of earlier samples ( 1 ) . That the test had factorial invariance for this sample and is homogeneous is apparent and may be considered as substantial evidence for its construct validity, as well as simple structure.