Statistical methods in rehabilitation research.

Of 128 articles in Volume 63 (1982) of the Archives of Physical Medicine and Rehabilitation 48 (37.5%) utilized some form of statistical methodology. The most commonly encountered methods were: mean (with or without standard deviation or standard error); analysis of variance (ANOVA); t-tests; simple correlation/linear regression; and chi-square analysis. A critical review of the analytic statistics used in 40 of these articles revealed that only 23 (57.5%) were considered satisfactory in descriptive and methodologic terms. The most frequently encountered deficiencies were: inadequate description of methods; misuse of t-tests; absence of appropriately complex statistical analysis. A greater understanding of and familiarity with biostatistics is essential for all readers of the Archives to allow critical evaluation of the rehabilitation literature.