Why researchers should not ignore measurement error and skewness in questionnaire item scores

Researchers commonly study associations between latent variables measured with items showing ordinal and skewed score distributions. Researchers can estimate associations between latent variables according to several statistical methods that vary for instance in how they handle the presence of skewness and measurement error in the item scores. In this commentary I use a computer simulation to illustrate that ignoring skewness or measurement error in questionnaire item scores often results in biased effect estimates, especially when testing interaction effects.