Comparison of methods for repeated measures binary data with missing values

• Author / Creator

  Mohammadi, Farhood

• Missing data is common in health and medical experiments including con- trolled clinical trials. Using only the complete cases for analysis may cause biased inferences or even controversial results. Assuming that the data are missing at random (MAR), various methods have been developed to handle missing data. Among them, GEE, non-linear mixed effects and multiple im- putation (MI-GEE and MI-NLME) methods based on GEEs and NLMEs are considered the most efficient methods. However, these guidelines are too lim- ited to apply generally. We evaluated their performance on various missing data mechanisms with repeated measurement binary data using a simulation study. We considered two different levels of correlation (ρ=0.3, 0.7) with three cases of repeated measures (T=2, 4, 6) with sample sizes of 40, 80, 200 under different missing data mechanisms, MCAR (Missing Completely at Random), MAR (Missing At Random) and MNAR (Missing Not At Random). Based on obtained empirical size control, power level and bias of each method, we conclude that the NLME based multiple imputation (MI-NLME) method performs the best.

• Subjects / Keywords

  • Missing values
  • Repeated measures

• Graduation date

  Fall 2014

• Type of Item

  Thesis

• Degree

  Master of Science

• DOI

  https://doi.org/10.7939/R3HT15

• License

  This thesis is made available by the University of Alberta Libraries with permission of the copyright owner solely for non-commercial purposes. This thesis, or any portion thereof, may not otherwise be copied or reproduced without the written consent of the copyright owner, except to the extent permitted by Canadian copyright law.