Resumen de Letters to the Editor: Authors’ Response

Alessandro Geminiani, Jack G. Caton, Carlo Ercoli, Changyong Feng

• Authors’ Response:

  We are happy to respond to the comments by Drs. Park, Ko, and Park related to the statistical methods and assumptions used in our study. The following points address their concerns:

  1. Defining a study population depends on the purpose of a study. No matter whether the data are from a census or from a sample survey, they can always be embedded in a “super” sample. Alternatively, the collected data can be regarded as a population. This is easily done by defining an appropriate probability space. Once the probability space is defined, statistical methods can be used to analyze the data. Please refer to Durrett1 and Shao2 (chapter 1) for more details.

  2. A statistical model is an approximation to the distribution of the population (which is never known) we want to study. If the sample size is large enough, it doesn’t matter whether n or n-1 is used in the denominator in estimating the standard deviation.

  3. It seems that Dr. Park et al. are trying to challenge the meta-analysis. Meta-analysis is a rigorous statistical method based on a solid theoretical background. Of course, it has its own shortcomings. For example, selection bias may void the analysis results. However, this is totally out of the control of the meta-analysis. See Whitehead3 for more details.

  We hope this clarifies the questions posed by Drs. Park, Ko, and Park.

  The authors report no conflicts of interest related to this letter.