IN general, there are three ways of dealing with the problem of measurement error in path analyses of social and psychological data. Correction of parameter estimates for the attenuating effects of random error of measurement may be accomplished (1) by the use of a priori estimates of measurement error, (2) by designing into studies alternate measures of the same construct, or (3) by repeated measurements on the same population over time. This paper considers models for the estimation of measurement error in the latter case. Taking his lead from Coleman (1968), Heise (1969) has formulated a path analysis model for the assessment of reliability when a variable is observed at three or more points in time. A consequence of this model is a simple formula for reliability uncontaminated by instability in the true scores. One of the several assumptions necessary to the empirical validity of Heise's model, i.e., the assumption that the reliability of the measured scores is stable over time, is argued to be doubtful.
[1]
Malcolm E. Turner,et al.
The Regression Analysis of Causal Paths
,
1959
.
[2]
D. R. Heise.
Separating reliability and stability in test-retest correlation.
,
1969
.
[3]
O. Kempthorne,et al.
STATISTICS AND MATHEMATICS IN BIOLOGY
,
1955
.
[4]
Hubert M. Blalock,et al.
Causal Inferences, Closed Populations, and Measures of Association
,
1967,
American Political Science Review.
[5]
A. Portes,et al.
The educational and early occupational attainment process.
,
1969
.
[6]
O. D. Duncan.
Path Analysis: Sociological Examples
,
1966,
American Journal of Sociology.
[7]
O. D. Duncan,et al.
The American Occupational Structure
,
1967
.
[8]
M. R. Novick,et al.
Statistical Theories of Mental Test Scores.
,
1971
.
[9]
J. B. Pearson,et al.
Methodology in Social Research.
,
1968
.