Inference from Nonrandomly Missing Categorical Data: An Example from a Genetic Study on Turner's Syndrome

Abstract

The process leading to partial classification with categorical data is sometimes nonrandom. A particular model accounting for incomplete data, which allows the probability of uncertain classification to depend on category identity, is utilized for an analysis of data obtained from a genetic study on Turner's syndrome. Estimates of population proportions are obtained from maximum likelihood. A method for handling nonrandomly missing data arrayed in contingency tables is discussed. Sensitivity analyses incorporating parameters related to the missing-data mechanism are recommended for estimation and testing.

Key Words:

• Contingency tables
• Missing data
• Non-randomness
• Partial classification
• Sensitivity analysis
• Turner's syndrome