The "power" of sound statistics

Article Abstract:

As a biostatistical consultant to The Journal of the American Medical Association, the author has reviewed manuscripts submitted for publication. The lack of adequate justification for sample size is a major statistical limitation of many of these articles. A greater awareness of the importance of the proper application of statistics to medical research is needed. The article by Arkin and Wachtel that appears in the January 12, 1990 issue of this same journal provides a good outline of basic principles for medical research. Although the article reveals nothing new in the field of statistics, a straightforward and simple approach to methods of test assessment and screening is presented. Two points that are mentioned deserve special emphasis. The methods described for the determination of sample size were not intended to be rigid, but to "indicate the orders of magnitude". In the author's personal experience sample size most effectively serves as a starting point and the size is ultimately determined by a combination of factors. The second point the author stresses is treatment of the research study that reaches negative conclusions. For this type of study, and there are many, it is important that the statistical power be sufficient to detect any clinically relevant effects before reporting any meaningful conclusions. The author also notes and welcomes the improvement of communication between clinicians and biostatisticians in recent years. The use of sound statistical practices are valuable and can enhance the progress of medical research.

---

How many patients are necessary to assess test performance?

Article Abstract:

Questions regarding the clinical relevancy of medical research have recently received a great deal of attention. As a part of this, focus has been placed on sample size requirements in medical research. Statistical power analysis has also has played a significant role in the development of methods of comparison and quality control. An approach has been developed to determine the sample size requirement for test performance characteristics. Analysis of these characteristics allow assessment of the clinical usefulness of laboratory tests and provide a basis of comparison for different tests. Statistical comparisons are only useful if they can provide enough statistic power to make the sample size important in assessing performance characteristics. To avoid fewer tests with meaningless negative results and to allow a better quality of the evaluation of laboratory tests, tables have been designed to provide appropriate sample size requirements using standard formulas. Definitions of sensitivity, positive and negative predictive values, and Type I and II errors have been provided. Sample problems are given and the determination of meaningful confidence intervals are discussed. Assumptions must be stated that are inexact, but at the same time not completely arbitrary in determining proportional sample sizes. Estimates of observed proportions prior to the study and assignment of power and type I error may affect the accuracy of the calculated sample size.

Statistics, Diagnosis, Laboratory, Laboratory diagnosis

---

What's the Relative Risk? A method for correcting the odds ratio in cohort studies of common outcomes

Article Abstract:

The calculation of a risk or prevalence ratio in medical research can be statistically approximated from the odds ratio, avoiding complicated statistical techniques. While the odds ratio describes the occurrence of a disease or a treatment response in a study population, it only approximates the risk of the disease or response in some cases. The odds ratio indicates the probability of an outcome in one group, divided by the probability of an outcome in another group, as in a case-control experimental design. A simple calculation using the odds ratio and the outcome incidences can produce a corrected risk ratio.

Multivariate analysis, Health risk communication

Similar abstracts:

• Abstracts: The effect of the Medicare prospective payment system on the adoption of new technology: the case of cochlear implants
• Abstracts: Health objectives for the nation: progress toward achieving the 1990 objectives for the nation for sexually transmitted diseases
• Abstracts: A Southwest Oncology Group study on the use of a human tumor cloning assay for predicting response in patients with ovarian cancer
• Abstracts: Biliary perestroika. Biliary diversion for pancreatic carcinoma: matching the methods and the patient
• Abstracts: The multichain interleukin 2 receptor: a target for immunotherapy in lymphoma, autoimmune disorders, and organ allografts

This website is not affiliated with document authors or copyright owners. This page is provided for informational purposes only. Unintentional errors are possible.