There is no more striking fact in the history of international relations than that so serious a question should have been treated with such amateurish simplicity as has been the case in the plans for armament reduction." J.T. Shot well "Disarmament should be viewed as a process of creating conditions which make it difficult or impossible for nations to break or endanger the peace not as a bargaining process to reduce the cost of preparing for war." - Benjamin V. Cohen. The havoc caused by the two world wars convinced the people all over the world that if humanity is to be saved from complete annihilation, something should be done to reduce or limit certain or all types of armaments. Commenting on the need of disarmament Dr. Gerald Wendt' Observed; "If world war-III comes which we pray will never happen, it will be a war in which most people may die from silent, insidious anti-human weapons that make no sound, give no warning, destroy no forests or ships or cities but can wipe out human beings by the millions."

In this some new estimation methods and testing procedures for the linear regression models with heteroscedastic disturbances. A Minimum Norm Quadratic Unbiased (MINQU) estimation method has been developed for estimating the unknown heteroscedastic error variances by using the weighted studentized residuals. A multiplicative heteroscedastic linear regression model has been specified and a method of estimating the parameters of linear regression model along with the in the heteroscedastic error variance has been given by using the predicted residuals. Three types of modified estimators have been proposed for the parameter of multiplicative heteroscedastic error variance by using internally studentized residuals.an adaptive method of estimation has been suggested to estimate the heteroscedastic error variances based on Bartlett's test by using the internally studentized residuals. Besides these new estimation methods, the testing procedures for testing the equality between the regression coefficients in two/sets of linear regression models under heteroscedasticity have been suggested by using the studentized residuals.