Statistics Glossary → Parallel and strictly parallel measures model
Parallel and strictly parallel measures modelconcept
A classical test theory model that assumes all items in a scale measure the same true score with the same sensitivity and the same amount of error (and, in the strictly parallel model, also have equal means).
Classical test theory treats each item score as the sum of a true score and error, and defines a series of increasingly restrictive models for how items relate. In the congeneric model, item loadings and error variances are free. In the tau-equivalent model all loadings are equal while error variances may differ. In the parallel model both loadings and error variances are equal. The strictly parallel model adds the requirement that item means are equal as well.
These models can be fitted by maximum likelihood, in the manner of confirmatory factor analysis, and tested with a chi-square test; if the model fits, the common variance, true-score and error variances and the reliability can be estimated. Cronbach's alpha equals reliability only if tau-equivalence (or a more restrictive model) holds; with unequal loadings alpha understates reliability and McDonald's omega is more suitable. Strict parallelism rarely holds in real scales, so rejecting the model is not an alarm but a reason to move to a more flexible model.
An everyday example
A company weighs the same parcel on three different scales. If all three show the true weight on the same scale and each wobbles randomly by the same amount, the scales are 'parallel'. If, in addition, none of them weighs systematically heavy or light, so their average readings are also the same, they are 'strictly parallel'.
In MerQur
In Cronbach's Alpha and Reliability Analysis, the 'Additional output: Parallel / strictly parallel model' option fits both models by normal-theory maximum likelihood and reports the common variance, true-score and error variances, reliability and a chi-square test.
Analyses where this concept appears
Cronbach's Alpha and Reliability Analysis
Related concepts
McDonald's ω (omega) reliability coefficientSpearman-Brown formula (single and average measures)