[Missing data treatment method on cluster analysis]

by

Literature Review1

Criticism on Data Collection of Household Health Surveys1

Relative weight of responsiveness1

Domain weights1

Broader Health System and the Responsiveness1

Sources of information2

Translation, validity and reliability2

Universality of domains2

Non-users3

Suggestions for Analysing Survey Data3

Missing Data Treatment4

“Ad-hoc” Methods5

Multiple Imputation6

Conditional Gaussian7

Chained Equations7

Methods for Monotone Data sets8

Issues with Imputation9

Methods of Weighting9

Bayesian Approaches9

Cluster Analysis10

What is a cluster?10

Main Elements of Cluster Analysis12

Missing Data in Cluster Analysis12

Scope and Limitations18

Women Clusters Samples19

Interpretation20

Child Clusters Samples20

Interpretation20

Limitation21

Statement of the Problem21

Objective21

Importance22

Methodology23

Sudan Household Health Survey (SHHS)23

Sample Design23

Sampling frame and units of analysis24

Stratification24

Size and Allocation of Samples25

Sample selection procedures26

Estimation and weighting procedures29

Questionnaires29

Questionnaires Sample30

Data analysis31

Two-Step Cluster Analysis32

Assumptions of Data in Two-Step Cluster Analysis34

Two-Step Cluster Analysis Plots35

Two-Step Cluster Analysis Output35

References37

Literature Review

Criticism on Data Collection of Household Health Surveys

Household surveys performed by World Health Organisation (WHO) are often criticized for estimation of missing observations (Aalto 2000; Almeida et al. 2001; Williams 2000) and for the methodology for information collection (Blendon et al. 2001a; Navarro 2001; Almeida et al. 2001). Biasness in the information is also found in much criticism including the use of too small data for too many imputations, use of limited number of questions out of a large number of questions for indexing, inadequacy of the sample to represent the population, inherent flaws in method, and majority of key informants being the people of WHO (Williams 2005).

Relative weight of responsiveness

The framework of health surveys is often criticised by the commentators from regional consultations (WHO Regional Office for the Eastern Mediterranean 2001) of having responsiveness as it is relatively an important factor of household surveys.

Domain weights

In aggregating an overall responsiveness index, the relative weights of the seven domains were criticized by several commentators (WHO Regional Office for Africa 2001 WHO)

Broader Health System and the Responsiveness

Regional consultations (WHO Regional Office for South-East Asia 2001) have been always commenting the inability of survey responsiveness to represent the health system with its broader boundaries including the services of health promotion and protection like public access to information (Oswaldo Cruz Foundation 2000; Travassos 2001; Ugá et al. 2001).

Sources of information

Whether the users of a health system can better judge it or the key informants? This question has been arising by critiques (Blendon et al. 2001a). Canvassing the information from population crucially requires measure of responsiveness as well as satisfaction (Blendon et al. 2001b).

Translation, validity and reliability

Translation of the concept of cross-culture validity and responsiveness is described in the question raised by Aalto (2000). Issues raised by several regional consultations included the cross cultural validation of the household health surveys (WHO Regional Office for the Western Pacific 2001). In comparison with the modules owing to the abstractness of the involved concepts, translation responsiveness might be a slightly more difficult problem (Almeida et al. 200; WHO Regional Office for the Western Pacific 2001).

The responsiveness key informant instrument has been criticized by critiques and participants in regional consultations for the availability of standard instrument psychometric data (Aalto 2000; WHO Regional Office for South-East Asia 2001). Aalto (2000) and the regional consultations WPRO and SEARO indicated that any subsequent responsiveness questionnaire instruments ...