dc.contributor.author | Mueke, Austin Luki | |
dc.date.accessioned | 2013-11-25T12:14:04Z | |
dc.date.available | 2013-11-25T12:14:04Z | |
dc.date.issued | 2013-11 | |
dc.identifier.citation | Master of Science in Medical Statistics | en |
dc.identifier.uri | http://erepository.uonbi.ac.ke:8080/xmlui/handle/11295/60036 | |
dc.description | Presented in partial fulfillment for the requirements
For the degree of
Master of Science in Medical Statistics
University of Nairobi | en |
dc.description.abstract | This master research project compares the performance of Classification and Regression
Trees (CART) and Logistic Regression in studying determinants of pregnancy wastage using
pregnancy information from a population-based sample survey, The Kenya Demographic and
Health Survey 2008/2009.The project report also describes in detail the fundamental
principles of tree construction, splitting algorithms and pruning procedures. It also briefly
introduces the logistic regression and then shows the comparisons of the analysis results from
the two statistical methods using Receiver Operating Curve, Variable Importance and
Hosmer-LemeshowModel Goodness of Fit Tests. Logistic regression performed slightly
better than CART using AUC with both agreeing on age of the woman as the most important
determinant of pregnancy wastage. CART found that the age of the woman, highest level of
educational attainment, age at first birth, Type of place of residence being either ourban or
rural and birth order to be the most important determinants of pregnancy wastage. Logistic
regression analysis found out that Age of the woman, marriage to first birth interval, usage of
anti-malarial during pregnancy, type of place of residence and usage of iron supplementation
during pregnancy to be the most important determinants. The Hosmer-Lemeshow Goodness
of Fit Test showed that CART didn’t fit the well the data while the Hosmer-Lemeshow
Goodness of Fit Test for logistic regression showed that did fit the data well.The lack of close
fit for the data could be explained by the nature of data and this needs further investigation
comparing fits both population based data and obstetric data. However, CART results could
be used for selection of key variables to be used in logistic regression analysis. When applied
prudently, both CART and logistic regression are suitable for the analysis of the determinants
of pregnancy wastage. | en |
dc.language.iso | en | en |
dc.publisher | University of Nairobi | en |
dc.title | Use of CART and logistic regression analysis to identify key determinants of pregnancy wastage | en |
dc.type | Thesis | en |
dc.description.department | a
Department of Psychiatry, University of Nairobi, ; bDepartment of Mental Health, School of Medicine,
Moi University, Eldoret, Kenya | |
local.publisher | Institute of Tropical and Infectious Diseases (UNITID) | en |