ISSN: 1885-5857 Impact factor 2023 7.2
Vol. 69. Num. 12.
Pages 1235-1236 (December 2016)

Letter to the editor
Why Not Use Existing Knowledge: Bayesian Statistics. Response

Por qué no utilizar el conocimiento previo: la estadística bayesiana. Respuesta

Javier Aranceta-BartrinaabcdCarmen Pérez-RodrigobcNatalia Ramos-CarreraeSonia Lázaro-Masedoe
Rev Esp Cardiol. 2016;69:1234-510.1016/j.rec.2016.08.015
Daniel Hernández-Vaquero, Rocío Díaz, Jacobo Silva, César Morís

Options

To the Editor,

We would like to thank Hernández-Vaquero et al. for their interest and comments on our study.1 We agree that a Bayesian approach could enhance the analysis of data from the ENPE study (Spanish acronym for the Nutritional Study of the Spanish Population), and we will consider this for future publications. The debate on Bayesian vs frequentist methods has been open for some time.2,3

We used frequentist inference to analyze data collected from a random probability sample (n=3966), with a careful methodological protocol and quality controls. All the studies we used as reference and context, conducted in Spain and other countries, used this approach. Hernández-Vaquero et al. state that their Bayesian estimate coincides almost exactly with our frequentist estimate, which often happens when the studies are of similar design and the sample size is large.

We share the view of many other authors, that neither approach is superior: each has its advantages and limitations. It is true that interest in Bayesian methods is increasing, as reflected in the changes in the number of publications retrieved when searching the term “Bayesian” in PubMed.4 In studies from the last 6 years (2010-2015), 16 665 publications include “Bayesian” in the title or abstract, and 81 321 include “obesity”, but only 71 records contain both “Bayesian” and “obesity”. Most epidemiological research has been done (and continues to be done) using a frequentist approach, without jeopardizing the knowledge acquired. Many authors use both approaches, depending on the research question, the study design, the size and design of the sample, the type of data, etc.5 We advocate a pragmatic approach, based on reasoning, reflection, and contextualization of the data.

FUNDING

The ENPE study was funded by the Fundación Eroski via an agreement with SPRIM and the Spanish Society of Community Nutrition (SENC). The sponsor was not involved in the study design, data collection, analysis or interpretation of results, drafting of the manuscript, or the decision to publish the results.

CONFLICTS OF INTEREST

S. Lázaro-Masedo and N. Ramos-Carrera are linked to SPRIM, who carried out consulting activities for the Fundación Eroski.

References
[1]
J. Aranceta-Bartrina, C. Pérez-Rodrigo, G. Alberdi-Aresti, N. Ramos-Carrera, S. Lázaro-Masedo.
Prevalencia de obesidad general y obesidad abdominal en la población adulta española (25-64 años) 2014-2015: estudio ENPE.
Rev Esp Cardiol., (2016), 69 pp. 579-587
[2]
L.C. Silva, A. Muñoz.
Debate sobre métodos frecuentistas vs bayesianos.
Gac Sanit., (2000), 14 pp. 482-494
[3]
S. Greenland.
Bayesian perspectives for epidemiological research: I. Foundations and basic methods.
Int J Epidemiol., (2006), 35 pp. 765-775
[4]
E. Zangiacomi-Martinez, J. Alberto-Achcar.
Trends in epidemiology in the 21st century: time to adopt Bayesian methods.
Cad Saude Publica., (2014), 30 pp. 703-714
[5]
L. Seliske, T.A. Norwood, J.R. McLaughlin, S. Wang, C. Palleschi, E. Holowaty.
Estimating micro area behavioural risk factor prevalence from large population based surveys: a full Bayesian approach.
BMC Public Health., (2016), 16 pp. 478
Copyright © 2016. Sociedad Española de Cardiología
Are you a healthcare professional authorized to prescribe or dispense medications?