Robust confirmatory factor analysis based on the forward search algorithm (Q2442683)

The paper uses the forward search algorithm within confirmatory factor analysis in order to achieve an ordering of the data based on residuals. The methodology starts from a basic set and observations are iteratively introduced by monitoring several estimates and indices. Once the algorithm is finished, several basic sets (each from a step of the iterative process) are available and the forward plots are next inspected in order to identify the observations that are influential on the model parameters and fit indices. The proposed approach thus proves to be effective by achieving an ordering of the influence of each observation based on its degree of closeness to the model and by additionally providing forward plots, which are informative of the dynamics of the entire process.