Statistical models are also needed for another reason. If we measure happiness with psychometric measures, we have to deal with another source of change: measurement error. Measurement error cannot be avoided in the empirical sciences. In contrast to state change and trait change, it is a form of unsystematic change. If we do not correct for measurement error, systematic change might be overestimated. Moreover, measurement error can produce some statistical artifacts in longitudinal data analysis such as the regression toward the mean and an artificial correlation between change and baseline scores (Rogosa, 1995). Consequently, statistical models are needed that are able to separate three types of change: Unsystematic variability that is due to measurement error , Systematic state variability that is due to situational influences and the interaction between the person and the situation, Systematic trait change that can be due to events, maturation, and other reasons