Mohamad, Ismail and Harun, Sobri and Abdul Malek, Marlinda (2006) Missing rain fall data imputation via expectation maximization (EM) algorithm. In: Recent Advances in Probability and Statistics. Penerbit UTM, Johor , pp. 67-90. ISBN 978-983-52-0612-2
Full text not available from this repository.
Official URL: http://www.penerbit.utm.my/bookchapterdoc/FS/bookc...
Missing data is a common problem encountered by many practicing statisticians and engineers. The daily rainfall data in Malaysia known as National Network (“NN”) System is no exception. Missing data generates gaps in the data matrix which posed problems in applying standard statistical methods to such data matrix. Thus, it is natural to replace the gaps with a probable value with the intention of producing completed data matrix. In Malaysia, three methods of recording daily rainfall data are used, namely manual, chart and data logger methods. Some of these recordings may be missing for a particular day which generate incomplete unit. When all the three recordings are missing for a particular day it generates unit missing. The current practice is to replace the missing data with the available values when incomplete unit is encountered. This technique is called imputation in statistical literature. However, the present method fails to address the unit missing problem. This study introduces an alternative statistical model which could cope with unit missing problem. The focus of this study is on Expectation Maximization (EM) Algorithm procedure in imputing missing daily rainfall values when dealing with incomplete units and Nearest Neighbor (NNeigh) Imputation technique when dealing with unit missing.
|Item Type:||Book Section|
|Deposited By:||Liza Porijo|
|Deposited On:||25 May 2012 04:12|
|Last Modified:||05 Feb 2017 03:38|
Repository Staff Only: item control page