Research in physical medicine and rehabilitation. IX. Primary data analysis

The primary statistical analysis is approached from the standpoint of what is required to publish in medical research journals. Descriptive and bivariate statistics cover the majority of medical research articles now published. Statistical guidelines for review of manuscripts are used to develop guidelines for analysis, including specifying the objective, the source of subjects and response rate, differences detectable with the expected sample size, appropriateness of statistics for one and two variables, method of presentation of results and conclusions and calculation of confidence intervals. Common mistakes to avoid include use of standard error of the mean instead of standard deviation, use of standard deviation with skewed data, failure to describe the statistical test used, multiple comparisons and failure to use special forms of t test and chi 2.