Yu-Kang Tu1, Katherine Keyes2, George Davey Smith3. 1. Department of Public Health, Institute of Epidemiology and Preventive Medicine, College of Public Health, National Taiwan University, Taipei, Taiwan. Electronic address: yukangtu@ntu.edu.tw. 2. Department of Epidemiology, Columbia University, New York, NY. 3. MRC Integrative Epidemiology Unit (IEU), University of Bristol, Bristol, UK.
Abstract
PURPOSE: Identification is a central problem with age-period-cohort analysis. Because age + cohort = period, there is no unique solution to the linear effect using generalized linear modeling, but cohort effects have caused greater controversy than age and period effects. To illustrate the magnitude of cohort effects given the presence of collinearity, we reanalyze data from the seminal study by Kermack et al, with an update. METHODS: Relative mortality data in England and Wales between year 1845 and 1995 were analyzed using partial least squares regression. There were seven age groups ranging from 5 to 74 years old and 16 periods with 22 cohorts. RESULTS: Our reanalysis seemed to support the existence of cohort effects in the mortality trends. Period and cohort effects were generally consistent with changes in the social, economic, and environmental factors taking place in the last two centuries. Our analysis also showed a declining trend in period effects up to 1950s. CONCLUSIONS: Partial least squares and related methods provide intuitive pointers toward the separation of linear age, period, and cohort effects. Because statistical algorithms cannot distinguish between relative and actual mortality rates, cohort effects may be underestimated because of contamination by negative age effects.
PURPOSE: Identification is a central problem with age-period-cohort analysis. Because age + cohort = period, there is no unique solution to the linear effect using generalized linear modeling, but cohort effects have caused greater controversy than age and period effects. To illustrate the magnitude of cohort effects given the presence of collinearity, we reanalyze data from the seminal study by Kermack et al, with an update. METHODS: Relative mortality data in England and Wales between year 1845 and 1995 were analyzed using partial least squares regression. There were seven age groups ranging from 5 to 74 years old and 16 periods with 22 cohorts. RESULTS: Our reanalysis seemed to support the existence of cohort effects in the mortality trends. Period and cohort effects were generally consistent with changes in the social, economic, and environmental factors taking place in the last two centuries. Our analysis also showed a declining trend in period effects up to 1950s. CONCLUSIONS: Partial least squares and related methods provide intuitive pointers toward the separation of linear age, period, and cohort effects. Because statistical algorithms cannot distinguish between relative and actual mortality rates, cohort effects may be underestimated because of contamination by negative age effects.
Authors: Katy Tobin; Mark S Gilthorpe; James Rooney; Mark Heverin; Alice Vajda; Anthony Staines; Orla Hardiman Journal: J Neurol Date: 2016-07-02 Impact factor: 4.849