An improved double sampling procedure based on the variance

Area of Science:

• Biostatistics
• Statistical Methods
• Clinical Trial Design

Background:

• Accurate sample size calculation is vital for study power, preventing underpowered or overpowered research.
• Adaptive sample size methods, using subsample variance for recalculation, are gaining popularity.
• Traditional methods like Stein's procedure are underutilized due to perceived data exclusion.

Purpose of the Study:

• To address limitations of existing adaptive sample size methods.
• To develop an unbiased variance estimation technique utilizing all available data.
• To ensure the proposed method maintains a controlled Type I error rate.

Main Methods:

• Application of the Helmert transformation to analyze variance estimation.
• Development of a novel unbiased variance estimator incorporating all study data.
• Mathematical proof demonstrating the Type I error rate does not exceed alpha.

Main Results:

• The naive approach using all data for variance estimation leads to underestimation of the true variance.
• The proposed method provides an unbiased variance estimate using the complete dataset.
• The Type I error rate of the developed procedure is rigorously proven to be controlled.

Conclusions:

• The proposed unbiased variance estimation method offers a statistically sound approach for adaptive sample size calculations.
• This method overcomes the limitations of previous techniques by utilizing all data effectively.
• The guaranteed control of the Type I error rate enhances the reliability of studies employing this adaptive strategy.