The Use of Friedman's Test in the Case of Informatively Missing Data

• Activity Number: 238
• Type: Contributed
• Date/Time: Monday, August 4, 2014 : 2:00 PM to 3:50 PM
• Sponsor: Section on Nonparametric Statistics
• Author(s): Annie Howard*+
• Companies: University of North Carolina at Chapel Hill
• Keywords: Randomized Block Design ; Missing Data

Abstract

Many public health studies utilize a randomized block design, where the block group is not of primary interest to the investigators but analysis should account for the intra-block correlation that it introduces. Friedman's test is one of the most common methods of analysis in this type of setting when one is not willing to make distributional assumptions about the data points. Friedman's test requires complete and balanced data; however, researchers have developed methods to adapt Friedman's test to situations involving missing completely at random data. Often, the reason for the data to be missing is directly related to the outcome values (for example, the case of detection limits), and the assumption that the data are missing completely at random is too strict an assumption. A new strategy is proposed for adjusting Friedman's test to informative missing data scenarios. The method put forth in this paper involves single imputation to impute missing rank values along with an optional weight, which gives less weight to individuals with more missing data, when there is some uncertainty regarding the missing data mechanism. Guidelines and suggestions, in particular for using this met