Related Experiment Videos
Confidence intervals in medical research.
S Radhakrishna1, B N Murthy, N G Nair
1Institute for Research in Medical Statistics, Madras.
The Indian Journal of Medical Research
|June 1, 1992
Summary
Confidence intervals offer more medical insights than significance tests. The Square Root Transformation method is recommended for binomial proportions, and Jeffreys-Perks for differences between small sample proportions.
Area of Science:
- Medical Statistics
- Biostatistics
- Epidemiology
Background:
- Confidence intervals (CIs) are crucial in medical research, offering more information than simple significance testing.
- CIs are widely applicable in clinical trials, disease control, vaccine studies, and laboratory research.
- Accurate calculation of CIs for binomial proportions can be computationally intensive.
Purpose of the Study:
- To re-emphasize the utility of confidence intervals in medical applications.
- To assess approximate methods for calculating confidence limits for binomial proportions.
- To recommend appropriate methods for calculating confidence intervals for single proportions and differences between two proportions.
Main Methods:
- Evaluation of 15 published methods for approximate confidence limits of binomial proportions.
- Comparison of accuracy and computational ease of different CI methods.
- Assessment of methods for confidence intervals of the difference between two proportions.
Main Results:
- The 'Square Root Transformation' method is recommended for calculating approximate confidence limits for binomial proportions due to its accuracy and ease of computation.
- For sample sizes exceeding 75, the usual method is acceptable for the difference between two proportions.
- For smaller sample sizes (even n=5), the Jeffreys-Perks method is highly satisfactory for the difference between two proportions.
Conclusions:
- Confidence intervals are more informative than statistical significance tests and should be used adjunctively in medical research.
- The 'Square Root Transformation' method provides an accurate and computationally feasible approach for binomial proportion confidence intervals.
- The Jeffreys-Perks method is recommended for calculating confidence intervals for the difference between two proportions, especially with small sample sizes.