Нестрогий байесовский подход к проверке допущения об измерительной инвариантности: иллюстрация на примере ценностей выбора
Measurement invariance is a key prerequisite for meaningful comparative studies using survey data. For many popular constructs in social sciences, unfortunately, measurement invariance does not hold. This paper illustrates a novel and efficient approach to measurement invariance testing, known as approximate Bayesian invariance. This approach has been shown in a couple of recent studies to be quite effective in handling a relatively small amount of non-invariance in latent constructs. This powerful methodology is applied to establishing cross-national comparability of pro-choice values, a sub-dimension of a well-known index of emancipative values (EVI) reflecting people’s attitudes to abortion, divorce, and homosexuality. Although the classical, or exact, approach to invariance testing suggests that neither full scalar, nor even full metric invariance hold for pro-choice values across ten cultural zones of the Welzel-Inglehart cultural map of the world, the Bayesian approach offers a rather optimistic conclusion that pro-choice values are approximately invariant not only across zones, but, for each World Values Survey wave separately, also across all countries covered in that wave.