Mittausmallin

The mittausmallin, in English usually referred to as the measurement model, is a component of structural equation modeling used to define how latent constructs are measured by observed indicators. In a measurement model, latent variables are hypothetical constructs such as intelligence, attitude, or satisfaction, which cannot be observed directly, while observed variables are indicators such as questionnaire items or test scores used to measure them. The model specifies the relationships between each latent construct and its indicators, typically through factor loadings and error terms. There are two main forms: reflective measurement models, where indicators are caused by the latent variable and share common variance, and formative measurement models, where the latent construct is formed by the indicators. In confirmatory factor analysis, the measurement model is specified a priori and tested for fit against data. Good fit indicates that the observed data are compatible with the hypothesized structure; poor fit may suggest misspecification, measurement error, or the need to revise indicators.

Reliability and validity assessments are integral: composite reliability and Cronbach's alpha assess internal consistency; convergent validity is often evaluated via average variance extracted; discriminant validity checks ensure construct distinctness. In practice, the measurement model precedes the analysis of structural relations among latent variables, and issues such as measurement invariance across groups are examined to support meaningful comparisons. Estimation methods include maximum likelihood and Bayesian approaches, with identification requiring fixed or constrained scales. The term mittausmallin is the Finnish genitive form of mittausmalli and is used in Finnish texts to refer to the measurement model, reflecting the broader aim of capturing the measurement properties of constructs through observed data in social science, psychology, and related fields.