Letters to the Editor: Re: Bibliometrics Study on Authorship Trends in Periodontal Literature From 1995 to 2010
- Autores: Jun-Beom Park, Youngkyung Ko
- Localización: Journal of periodontology, ISSN 0022-3492, Vol. 86, Nº. 1, 2015, págs. 7-7
- Idioma: inglés
- Enlaces
-
Resumen
To the Editor:
Re: Bibliometrics Study on Authorship Trends in Periodontal Literature From 1995 to 2010. Geminiani A, Ercoli C, Feng C, Caton JG. (J Periodontol 2014;85:e136-e143).
We have read the article by Geminiani et al. with great interest. Five journals with the highest impact factors—as recorded at the time of this review (the impact factor was recorded as of January 2011, single year)—that also represent the specialty and the periodontal organizations were selected: The International Journal of Periodontics & Restorative Dentistry, Journal of Clinical Periodontology, Journal of Periodontal Research, Journal of Periodontology, and Periodontology 2000. Two thousand seventy-six (2,076) articles were selected from a total of 2,260 articles that met the inclusion/exclusion criteria and were published in the years 1995, 2000, 2005, or 2010. Changes in the number of authors, study design, topic investigated, financial support, and geographic origin were analyzed. A general linear model (two-way analysis of variance with a second-order interaction) and chi-square tests were used for the statistical analysis, and the P value was applied to determine statistical significance.
We would like to raise several issues related to this article. The definition of a statistical test is a process to derive a generalized conclusion for the population using partial information (with possible uncertainty) from the sample. A population is the entire target that the researcher wants to know, regardless of its size of the number, while a sample means part of the population. Thus, a statistical test should be used only when the data obtained by the researcher are from a sample; this means that a statistical test must be used when a population (the target of the generalization) exists but is unknown.
The investigation of a population itself is called a census, while investigating the sample is called a sample survey. Thus, data from the census, which has information about the entire target of interest, no longer require interference—the data only need to be described. For example, if a population consists of 51 males and 49 females, it should be concluded that the number of males was two greater than the number of females. Thus, a statistical test should not be performed, and statistical significance or a P value of 0.05 must not be used.
Almost with no exception, many clinical research papers dealing with patients from one hospital present new findings using statistical analyses and P values. However, it can be questioned whether these data can be considered to be randomly extracted samples from a specific population. In reality, even though these are data collected during a specific period from one hospital, these can be considered to be sample data and statistical inference is generally applied, if a real or conceptual population can be described with well-defined inclusion/exclusion criteria.
However, in this study by Geminiani et al., all of the articles that met the inclusion/exclusion criteria were within the researchers’ interests and came from the five journals, and the four investigated time periods were research targets—they should be considered to be a population, not a sample. If this can be considered as a sample, then what should be the population itself? It seems sufficient to simply describe the trend within the investigated time points; statistical analysis and P values for trends should not have been mentioned in the article.
This confusion over the concept of the population and the sample in this study may have led to an error in calculating the standard deviations. A standard deviation is calculated using n−1 in the sample and n in the population. Thus, the true standard deviation may be smaller than the value presented in the article.
The authors report no conflicts of interest related to this letter.
