Application of a Simplex Process Model to Six Years of Cognitive Development in Four Demographic Groups

Abstract

A simplex process model of the cross-lagged correlation paradigm was applied to 16 tests administered to samples of black and white males and black and white females in Grades 5, 7, 9, and 11.Listening, a measure of aural comprehension, consistently anticipated individual differences on an intellectual composite in all four groups. The other achievement test of the STEP series anticipated individual differences on the so-called aptitude tests of SCAT, which in turn anticipated individual differences on the narrow information scores obtained from the Test of General Information (TGI). This model may be more powerful in revealing lags than the traditional methods of analyzing cross-lagged differences in longitudinal data. The model does not require stationarity and can produce a meaningful outcome in its absence.tions between true scores of X and Y on separate occasions to be equal to each other, it is neces-APPLIED PSYCHOLOGICAL MEASUREMENT

On the Measurement of Stability in Over‐Time Data

“…The equal stability assumption is only plausible when the set of measures IS taken from the same domain It would not be advisable to make that assumption, for example, if some measures are demographic and others are standard personality vanables Thus we did not include the interest vanables collected m the ETS data with the cognitive vanables because the stability of interests is lower (Humphreys, Park, & Parsons, 1979) While we do not have well-elaborated theones of which vanables are more stable, common sense and expenence will need to be exercised in choosing the vanables Campbell and O'Connell (1967) found strong evidence for equal stability across traits or factors So the assumption may not be as implausible as it might first seem…”

“…The two-component model is consistent with this pattern of results Time-senes methods measure the stability of the state factor because stability IS measured withm a person (lntra-individual change) If the state factor has low stability, then the time-senes estimates of stabihty should be low In the simplex model all unstable vanance is treated as error The simplex model measures the stability of the trait factor smce the state factor IS treated as error, so it should have the highest stability The twowave, multiple-vanable methods (whether one-or multiple-factor) tend to fall between time-senes and simplex estimates of stabihty For instance for the ETS data, simplex estimates of stability for the eight variables average 95 (Humphreys et al , 1979), which is higher than the two-wave, multiple-factor average of 92…”