Carel F W Peeters1, James Dziura2, Floryt van Wesel3. 1. Department of Epidemiology & Biostatistics, VU University Medical Center, Amsterdam, The Netherlands. Electronic address: cf.peeters@vumc.nl. 2. Yale Center for Analytical Sciences, Yale School of Public Health, New Haven, CN. 3. Department of Educational Neuroscience, VU University Amsterdam, Amsterdam, The Netherlands.
Abstract
PURPOSE: Factor analysis (FA) has become part and parcel in metabolic syndrome (MBS) research. Both exploration- and confirmation-driven factor analyzes are rampant. However, factor analytic results on MBS differ widely. A situation that is at least in part attributable to misapplication of FA. Here, our purpose was (i) to review factor analytic efforts in the study of MBS with emphasis on misusage of the FA model and (ii) to propose an alternative factor analytic strategy. METHODS: The proposed factor analytic strategy consists of four steps and confronts weaknesses in application of the FA model. At its heart lies the explicit separation of dimensionality and pattern selection and the direct evaluation of competing inequality-constrained loading patterns. A high-profile MBS data set with anthropometric measurements on overweight children and adolescents is reanalyzed using this strategy. RESULTS: The reanalysis implied a more parsimonious constellation of pathophysiological domains underlying phenotypic expressions of MBS than the original analysis (and many other analyses). The results emphasize correlated factors of impaired glucose metabolism and impaired lipid metabolism. CONCLUSIONS: Pathophysiological domains underlying phenotypic expressions of MBS included in the analysis are driven by multiple interrelated metabolic impairments. These findings indirectly point to the possible existence of a multifactorial etiology.
PURPOSE: Factor analysis (FA) has become part and parcel in metabolic syndrome (MBS) research. Both exploration- and confirmation-driven factor analyzes are rampant. However, factor analytic results on MBS differ widely. A situation that is at least in part attributable to misapplication of FA. Here, our purpose was (i) to review factor analytic efforts in the study of MBS with emphasis on misusage of the FA model and (ii) to propose an alternative factor analytic strategy. METHODS: The proposed factor analytic strategy consists of four steps and confronts weaknesses in application of the FA model. At its heart lies the explicit separation of dimensionality and pattern selection and the direct evaluation of competing inequality-constrained loading patterns. A high-profile MBS data set with anthropometric measurements on overweight children and adolescents is reanalyzed using this strategy. RESULTS: The reanalysis implied a more parsimonious constellation of pathophysiological domains underlying phenotypic expressions of MBS than the original analysis (and many other analyses). The results emphasize correlated factors of impaired glucose metabolism and impaired lipid metabolism. CONCLUSIONS: Pathophysiological domains underlying phenotypic expressions of MBS included in the analysis are driven by multiple interrelated metabolic impairments. These findings indirectly point to the possible existence of a multifactorial etiology.
Authors: Maria M Llabre; William Arguelles; Neil Schneiderman; Linda C Gallo; Martha L Daviglus; Earle C Chambers; Daniela Sotres-Alvarez; Diana A Chirinos; Gregory A Talavera; Sheila F Castaneda; Scott C Roesch; Gerardo Heiss Journal: Ann Epidemiol Date: 2015-02-19 Impact factor: 3.797