Corrigendum to “Community social networks, individual social participation and dietary behavior among older Japanese adults: Examining mediation using nonlinear structural equation models for three-wave longitudinal data” (Preventive Medicine (2021) 149, (S0091743521001973), (10.1016/j.ypmed.2021.106613))

Marisa Nishio, Daisuke Takagi, Tomohiro Shinozaki, Naoki Kondo

Research output: Contribution to journalComment/debate

Abstract

There was an error in one of the formulas used in the study. We deeply regret that such a mistake crept into the work, though it fortunately does not affect the overall findings. The error — the decimal point for the percentage of community social networks — was in the formula used to calculate the risk difference shown in Appendix B.1. We spotted the mistake when we were writing a report to the Japan Gerontological Evaluation Study, an organization that had provided us with the data used in this study. The error does not influence the overall findings because no results had been derived solely by referencing the risk differences. Specifically, the correction is irrelevant to the first finding, “individual social participation partially mediates the association between community social networks and fruit/vegetable intake”, because it stems from ‘a proportion mediated by the mediator’ and ‘probit regression coefficients’. The second finding, on the other hand, is impacted by the correction. The finding is “community social networks may encourage residents to participate in social activities” so that its result turned out to be underestimated due to the error. Nevertheless, its trend remains unchanged. The third finding, “the association between individuals' social participation and subsequent fruit/vegetable intake is greater among people living alone than among those living with someone”, is not pertinent to the correction as it does not relate to community social networks. The calculation of the risk difference using community social network as an explanatory variable was incorrect and under-estimated. The formula for calculating the risk difference shown in Appendix B.1 should have been as follows: [Formula presented] [Formula presented] The risk difference in Figs. 2 and 3, therefore, should have been reported as follows: Fig. 2 Path diagrams – stratification by household status.[Figure presented] Fig. 3 Path diagrams – stratification by household income.Notes: RD = risk difference, Digits beside the arrows = probit coefficients, and digits in parentheses = confidence intervales (CI) *Only the CI of study interests are shown. [Figure presented] Similarly, the risk difference in the Appendix D, E and F should have been reported as follows: Appendix D.1 [Figure presented] Appendix E.1 Path diagrams – stratification by sex.[Figure presented] Appendix E.2 Path diagrams – stratification by household status.[Figure presented] Appendix E.3 Path diagrams – stratification by household income.[Figure presented] Appendix F.1 Path diagrams – stratification by sex. [Figure presented] Appendix F.2 Path diagrams – stratification by household status.[Figure presented] Appendix F.3 Path diagrams – stratification by household income.[Figure presented]

Original languageEnglish
Article number106750
JournalPreventive Medicine
Volume153
DOIs
Publication statusPublished - Dec 2021

Fingerprint

Dive into the research topics of 'Corrigendum to “Community social networks, individual social participation and dietary behavior among older Japanese adults: Examining mediation using nonlinear structural equation models for three-wave longitudinal data” (Preventive Medicine (2021) 149, (S0091743521001973), (10.1016/j.ypmed.2021.106613))'. Together they form a unique fingerprint.

Cite this