A Two-Sample Nonparametric Test with Missing Observations

SYNOPTIC ABSTRACT

The Wilcoxon two sample nonparametric test is modified to analyze the data with missing observations. The approach is to weigh each complete observation with its own estimated weight score, which is the inverse of probability of completing the observation. The proposed modified method is applied to a recently conducted clinical trial data. A small simulation study shows that the modified method is unbiased under the null hypothesis, while the Wilcoxon two sample test is biased when applied only to complete observations.

Key Words and Phrases:

• Nonparametric Test
• Two Sample Test
• Missing Observations
• Propensity Score
• Clinical Trial