Related Experiment Videos
Non-parametric test of ordered alternatives in incomplete blocks
1Biometry and Mathematical Statistics Branch, National Institutes of Health, 6100 Executive Blvd, Room 7B13, Rockville, MD 20892, USA.
Statistics in Medicine
|May 18, 2000
Summary
This study introduces a new non-parametric method to detect time trends in medical studies with incomplete data. The approach uses linear regression and the Wilcoxon test, proving effective for analyzing trends in repeated observations over time.
Area of Science:
- Biostatistics
- Medical Data Analysis
Background:
- Medical studies often involve repeated observations over time, forming natural blocks of data.
- Missing observations can lead to incomplete data blocks, complicating trend analysis.
Purpose of the Study:
- To propose a novel non-parametric method for detecting ordered alternatives (time trends) in incomplete block designs.
- To provide a statistically sound approach for analyzing longitudinal data with missing values.
Main Methods:
- Estimating the linear trend within each incomplete block using linear regression.
- Applying the one-sample Wilcoxon test to the estimated linear trends to detect a significant trend.
- Assessing the asymptotic normality and consistency of the proposed test statistic.
Main Results:
- The proposed method effectively detects trends when they exist, demonstrating sensitivity.
- Monte Carlo simulations indicate competitive performance compared to extended Page and Jonckheere tests.
- The method allows for the estimation of the overall trend magnitude and its confidence interval.
Conclusions:
- The developed non-parametric test is a valuable tool for analyzing time trends in medical studies with incomplete longitudinal data.
- This method offers a robust alternative for situations where traditional parametric assumptions may not hold.