An alternative method to solve the transformation problem and its application to biostatistics
Loading...
Date
2009
Authors
W Ahmad, Wan Muhamad Amir
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
The aim of this research is to expose the issue of the data problems from various
studies which do not fulfill the assumptions for analysis of variance (ANOVA),
regression analysis, and other such procedures in detail together with the steps to
overcome this problem with modifying the original of Box-Cox formula until it can
be used for all the real numbers. The whole transformation analyses are emphasized
on achieving normality, variance stability and independence of observations and to
convert the model into a linear form. In this thesis, the SHAMIRIFF transformation
method is being introduced based on some modification of Box-Cox transformation
formula with some approaches to Optimization and Newton Raphson-Method. This
research is mainly focus on two methods which are normally used that are diagnostic
residual plots and the normal probability plot to measure the accuracy of the
transformation. A deeper study on transformation has been done by using data from
Box-Cox Transformation and also SHAMIRIFF transformation method. The residual
results from both methods are being compared relatively by using the value of
Anderson Darling (AD) statistics and P statistics. The transformation is carried out
for the response variables and the comparison results from both methods will provide
information on the success of the transformation. All the significant results will be
illustrated by using MAPLE and MINITAB software.
Description
PhD
Keywords
Chemical science , Transformation problem , Biostatistics