Possibilities and potentials of the structural equation modeling in analisys of between-group differences
Abstract
Diagnostics in sport psychology very often requires measurement of properties (theoretical constructs) that are not directly observable [1] and are therefore represented by manifest indicators. Sport motivation is just one example of such a construct.
The simple sum of the results in individual indicators serves in many cases as an estimate of the latent construct. Structural equation modeling (SEM) on the other hand offers a possibility to control the relative importance of indicators in the composite score.
Gender differences in intrinsic motivation (measured by The Sport Motivation Scale) were examined in a sample of 456 university students. Using an independent-group t-test, two types of sum score (simple sum vs.
SEM based sum) were compared. The difference between men and women in intrinsic motivation was not statistically significant for simple sum score (t(454) = 0,13; p = 0,90).
The difference between men and women in intrinsic motivation was statistically significant for SEM based sum score (t(454) = 2,38; p = 0,02). Since SEM offers a possibility to create methodologically more acceptable composite score than the simple sum of indicators, we may recommend SEM as a very useful tool in analysis of between-group differences especially in situations when measuring latent theoretical constructs.
Using SEM in these cases may have substantial impact on interpretation of our results.