Open AccessParametric and Nonparametric Methods in Medical Statistics
Author(s) -
А. Н. Герасимов,
Н. И. Морозова
Publication year - 2015
Publication title -
èpidemiologiâ i vakcinoprofilaktika
Language(s) - English
Resource type - Journals
eISSN - 2619-0494
pISSN - 2073-3046
DOI - 10.31631/2073-3046-2015-14-5-6-12
Subject(s) - nonparametric statistics , parametric statistics , statistics , kurtosis , ranking (information retrieval) , statistical hypothesis testing , computer science , econometrics , margin (machine learning) , mathematics , artificial intelligence , machine learning
The article is devoted to the conditions of applicability of parametric and nonparametric methods, including criticism of frequent methodological errors. By using methods of parametric statistics often make the wrong conclusion that to test the applicability of methods of parametric statistics need to find out whether there are significant differences resulting from the normal allocation of the pilot. Doing this is not necessary, as any encountered in biomedical research component is distributed clearly abnormal. Besides, for the practical application of parametric statistics need sufficient proximity to the source distribution is not normal, and the arithmetic mean from a set of observations. The article stated, under any circumstances and for any amount of observational techniques can be used parametric statistics and how it relates to the value of kurtosis, and what the margin of error calculation of significant differences when using methods of parametric statistics. Subjected to a critical analysis and methods of nonparametric statistics. With their more formal precision and breadth of applicability, however, they checked less valuable statistical hypothesis, as in fact do not work with the values, and their ranks, and the ranking much of the information is lost.