Estimating growth using the censored-inflated structural equation model with floor effects data

Abstract

A “floor effect” arises when performance is as bad as possible in all conditions. It is also known as the lowest possible measure of an individual’s performance or achievement. The floor may serve as a proxy for how low an observation could actually go if not for the lower bound which restricts the performance to a certain point. The floor effect can also be an artifact of an interaction between the assessment and examinees, which causes individuals to reach the bottom of their capacity to answer test items correctly. Much work has been conducted on factors affecting student achievement and achievement growth trajectories, however, there is very little regarding how to model the growth trajectories of a floor effect. The censored-inflated model is unique and seemingly appropriate to be employed with these types of data since two growth models are simultaneously estimated. The first is a continuous growth model, and the second a model specific for floor effects. A simulation study was conducted to test model fit under various conditions. The study provides promising evidence that the censored-inflated model is best used for these highly specific data.