Factor structural analysis of condition fluctuation by dynamic factor analysis

Abstract

The purpose of this study was to identify the factorial structure of perceived condition fluctuation considering a time lag in a single case design applying dynamic factor analysis (DFA). The subject was a female swimmer who had won the 200-m individual medley in the 1998 inter-collegiate competition in Japan. The 15 selected items measured her perceived physical and mental condition ; the amount of training per and her regular lifestyle were recorded by herself on the "Quality Control (QC) Sheet" every day for 167 days, 24 weeks. Common factors of condition fluctuation were extracted from multivariate time-series data using DFA with the maximum likelihood (ML) method and quartimax rotation. The lag-2 variance / covariance was most optimal for DFA according to autocorrelation and partial autocorrelation of all variable. Five factors and the lag-2 DFA model were adopted considering several model-fitting indicators. The factorial structure of condition fluctuation was different in each lag model, although such factors as the amount of practice, meal satisfaction, perceived condition, and fatigue were extracted in all lag models. It was inferred that the factorial structure of perceived condition fluctuation in a female swimmer could be constructed with a time lag consideration. It was expected that changes of factorial structure accompanying time progression could be confirmed by using DFA for multivariate time-series data.