Classroom observation rubrics are a widely adopted tool for measuring the quality of teaching and provide stable conceptualisations of teaching quality that facilitate empirical research. Here, we present four statistical approaches for analysing data from classroom observations: Factor analysis, Rasch modelling, latent class or profile analysis, and formative measurement models. Each statistical model conceptualises the latent variable differently, which may or may not align with the observation rubric's conceptualisation of teaching quality. We discuss the differences across these models, focusing on the alignment between the rubric's conceptualisation of teaching quality and the model's modelling of the latent variable. We discuss the need to align model selection with observation rubric so that the measured teaching quality reflects the theoretically conceptualised teaching quality.