Manuscript Title:

DEVELOPMENT OF MISSING DATA ESTIMATION TECHNIQUES WITH STATISTICS METHODS

Author:

Orathai Chuacharoen, Maniratt Jaroongdaechakul and Montree Piriyakul

DOI Number:

DOI:10.17605/OSF.IO/FC4JZ

Published : 2021-06-23

About the author(s)

1. Orathai Chuacharoen - Department of Statistics, Faculty of Science, Ramkhamhaeng University, Thailand.
2. Maniratt Jaroongdaechakul - Department of Statistics, Faculty of Science, Ramkhamhaeng University, Thailand.
3. Montree Piriyakul - Department of Statistics, Faculty of Science, Ramkhamhaeng University, Thailand.

Full Text : PDF

Abstract

Development of missing data estimation techniques with statistics methods is an experimental research in order to study solution process for missing data and develop new data analysis techniques and statistics estimation methods by using data with uniform distribution (0,1), binomial distribution (50, 0.2), binomial distribution (50, 0.5), binomial distribution (50, 0.8), normal distribution (0.1), and 1,000 sets of real data with 10,000 replications. Research results are shown as follows: 1. Mean Imputation (MI) is suitable for normal distribution data. 2. K-Nearest Neighbor Imputation (KNN) is suitable for normal distribution and real data when the sample size is large. 3. Extreme Imputation (EI) is suitable for discrete random variables especially in uniform and binomial distribution, MSE and MMRE will be low when the sample size is small. 4. Side Imputation (SI) is suitable for binomial distribution data. Proportion of missing data is high and the sample size is large. 5. Side Mean Imputation (SMI) is suitable for normal distribution and real data when the sample size is large. 6. New Multiple Imputation (Multi-Im) is suitable for binomial distribution data with more than 10 sample sizes. Concept of mean value application basis is used in study and development of missing data estimation technique. For further studies, concepts from other theories might be applied.


Keywords

Missing Data, Missing Data Estimation