EM or DA or EM and DA
Author: Chief Assist. Deyan Lazarov (Burgas Free University)
Keywords: EM Algorithm, Data Augmentation (DA), Missing values, K-means clusters, Labour Force Survey
The research extends the previous work of the author and presents the missing values (MV) problem in Labour Force Survey conducted in Bulgaria 2007 by National Statistical Institute. The mechanism of MV is MNAR and after K-means clustering procedure the database is split in to two general clusters Cluster 7 and the rest data. Missing data analysis is performed to two general clusters separately by using EM Algorithm, Data Augmentation Algorithm (DA) and the form of the DA with starting conditions based on EM initial parameters. The three methods are compared and specific conclusions are present.
