Title : Derivation Some Important Properties of Parameters for a Specific Form of Heteroscedastic Model and the Associated Tests

Authors : Md. Sharif Hossain

Abstract : The principal purpose of this thesis is to derive some important properties of the ordinary least squares estimators of parameters for a specific form of heteroscedasticity that is error variances are proportional to the square of the independent variable. Moreover, the particular form of heteroscedasticity has been considered for a specific linear regression equation without the intercept term imposing the additional condition that sum of squares of the observations of the independent variable is equal to the number of observations. Attempt has also been made to show theoretically whether GLSE of parameters are as efficient as OLSE for the assumed form of heteroscedasticity with the aditional condition. Another attempt that has also been made in the study to identify which tests are appropriate for detecting the presence of heteroscedasticity of the assumed form for a given data set. A modified new test following the work of Park (1966) for detecting the presence of heteroscedasticity of the assumed form has been proposed and the important properties of the proposed test statistic have also been derived. For emperical verification of the theoretical constructs of this study a set of real data on the total collection and collection of bound journals for 45 libraries in Bangladesh for the year 1995 has been used. The rationale behind the choice of the data set for the current purpose is also delineated. Using the real data set an attempt has also been made to select appropriate test procedure for detecting the presence of heteroscedasticity of the particular form on power property.

Journal : Volume : Year : 1995 Issue :
Pages : 166 City : Edition : Editors :
Publisher : ISBN : Book : Chapter :
Proceeding Title : Institution : University of Dhaka Issuer : Number :