[Evidence-based statistics: appropriate approach for description and analysis of research data]

ABSTRACT

In recent years using statistics in medical sciences is increasing and statistics is an essential part of each research. At present, for each study design, and for each type of data, there are appropriate statistical methods, which yield to valid results and inferences. Familiarity with such methods is essential, especially for those involved in the field of medical research, as they might confuse majority of readers who are not familiar with details of statistical analysis and data interpretation. Other consequences include waste of money, time, and energy spent for each research project. In recent years some papers have been published to enhance researchers' knowledge of statistics and its application in research methodology, but still there are many methodological mistakes in manuscripts. In this article we aim to demonstrate appropriate methods for data analysis. We will discuss importance of confidence intervals and their advantages over P-value. We will finally highlight common errors in data analysis such as univariate methods, ignoring power in detection of difference between groups, and using correlation coefficient to assess the agreement between scores