A practical approach for comparing means of two groups without equal variance assumption.

Area of Science:

• Biostatistics
• Statistical Inference

Background:

• Compares two independent groups of normally distributed random variables.
• Does not assume equal variances between the groups.

Purpose of the Study:

• Proposes a simple method for constructing confidence bounds using Howe's approximation I.
• Applies the method to parallel clinical trials and bioequivalence trials.
• Derives sample size calculation formulas for these trial types.

Main Methods:

• Utilizes Howe's approximation I for confidence bound construction.
• Applies the method to hypothesis testing in clinical trials (H0: mu(x)-mu(y)=0 vs H1: mu(x)-mu(y)<0).
• Applies the method to bioequivalence trials (H0: |mu(x)-mu(y)|>=delta vs H1: |mu(x)-mu(y)|

Main Results:

• The proposed method controls type I error satisfactorily.
• Maintains relatively high statistical power compared to existing methods.
• Offers a simple and accurate formula for sample size calculation.

Conclusions:

• The proposed procedure provides a simple and effective approach for confidence bounds.
• It is suitable for parallel clinical and bioequivalence trials with unequal variances.
• The derived sample size formulas are accurate and practical.