AUTHOR: NWANKWO CHIKE HENRY
AFFILIATION: NNAMDI AZIKIWE UNIVERSITY, AWKA
In this work, specially designed cumulatively coded dummy variable multiple regression models of temperature and precipitation were developed using the method of Path Analysis. We also developed absolute direct and indirect effects of each of the parent independent variables on the dependent variables through the mediation of their respective dummy variables in the presence of the ordinal dummy variables representing other parent independent variables in the models. The effects are estimated as the weighted averages of the effects of these ordinal dummy variables for each parent independent variable which are used in the fitted dummy variable multiple regression models. The weights which sum up to 1 are the estimated simple regression coefficients of the representative dummy variables on its parent variable. The models developed are applied to analyse temperature and precipitation for a 30 year data with temperature and precipitation as dependent variables, while using year, and month as parent independent variables, each of which is represented by a set of ordinal dummy variables in the regression models. These effects are used in the fitted models to predict temperature and precipitation for any desired parent independent variable. Although both month and year are found to have significant effects on temperature, only month has statistically significant effect on precipitation. Our dummy variable fitted model indicates a high level of association between the observed and the expected temperature values. Predicted values for May 2008 and September 2009, say, are 32oC and 30oC respectively with observed values of 32oC and 31OC, showing an error rate of 0oC and 1OC respectively. Error analysis show very low variation, indicating a good fit.
TO VIEW THE FULL CONTENT OF THIS DOCUMENT, PLEASE VISIT THE UNIZIK LIBRARY WEBSITE USING THIS LINK, http://naulibrary.org/dglibrary/admin/book_directory/Thesis/11081.pdf