Application of Linear Mixed-Effect Modeling for the Analysis of Human-in-the-Loop Simulation Experiments

Abstract

A real-time human-in-the-loop (HITL) simulation is a crucial validation activity in the life cycle of air traffic management operational concept development. When planning HITL simulations, however, researchers face a series of experimental design constraints that often limit the application of advanced statistical data analyses. Linear mixed-effects modeling (LMEM) is a multiple regression analysis technique that is comparatively flexible in considering the covariance present in repeated measures data. This paper aims to make LMEM better known to applied researchers in the field of aviation research, particularly those applying HITL. For this purpose, the building steps of LMEM, the output, and model-fit tests are explained based on data obtained from an Airbus 320 flight simulation study that examined the impact of two different wake separation schemes on either a final approach or a departure path on the pilots’ perceived severity of a wake-vortex impact. The experimental setup involved six experimental factors that were not fully crossed, had an unbalanced number of repeats per pilot, and contained missing data. This prevented the use of the more traditional repeated measures analysis of variance. The LMEM was able to handle this and could explicitly test the study hypotheses with statistical confidence.