stochasticrelationshipsandscatterdiagrams.ppt

社會統計（上） Stochastic Relationships and Scatter Diagrams 如果x與y的關聯分佈圖呈現一近似直線的關係，則我們假設兩者具有以下的線性關係： Stochastic Relationships and Scatter Diagrams Deterministic Relationship and Stochastic Relationships 我們的任務在於估計能夠正確描述X與Y關係的截距?0與斜率?1。 Deterministic Relationship and Stochastic Relationships Deterministic Relationship and Stochastic Relationships Deterministic Relationship and Stochastic Relationships Deterministic Relationship and Stochastic Relationships Deterministic Relationship and Stochastic Relationships Sample Regression Line 簡單迴歸模型的基本假設 Method of Least Squares 如何找出一條最具有代表性的回歸線？ 每一個人用肉眼判斷，得到的結果都不會一樣，因此需要一個較精準的方法。 Method of Least Squares 尋求迴歸係數的估計式有許多種方法，最常用的為 普通最小平方法(ordinary least squares method)及最大概似法(Maximum likelihood method) Method of Least Squares 由於母體參數β0,β1為未知數，因此母體迴歸線必須透過觀察到的樣本(xi, yi)來推估 尋求迴歸係數的估計式有許多種方法，最常用的為 普通最小平方法(ordinary least squares method)及最大概似法(Maximum likelihood method) Method of Least Squares 找出最適合樣本資料的一條直線，使預測值與觀察值的差異最小。 Method of Least Squares Residual Sum of Squares 由於殘差值有正有負，無法正確測量出兩者的距離，因此將殘差值加以平方。 我們的目標是要找出一條線，使每一個觀察值與預測值的距離的平方和最小： Residual Sum of Squares Stochastic Relationships and Scatter Diagrams 直線上任兩點P1P2，從P1移至P2，x軸座標移動 △x = x2 - x1 y軸座標移動 △y = y2 - y1 Stochastic Relationships and Scatter Diagrams 直線上任兩點P1P2，此線的斜率定義為： Stochastic Relationships and Scatter Diagrams 在直線上任意點（P1 或P2）的斜率皆相同，變遷速率為常數 微分(derivative)簡介 微分(derivative)簡介 微分(derivative)簡介 微分(derivative)簡介 微分(derivative)簡介 微分(derivative)簡介 微分(derivative)簡介 微分(derivative)簡介 Slope of the Tangent Line Slope of the Tangent Line Slope of the Tangent Line Slope of the Tangent Line Slope of the Tangent Line Derivative The derivative of function f with respect to x is the function f defined by Notation for the derivative f (x) 讀做? f prime of x y 讀做 y prime the derivative of y with respect to x  dee y dee x the derivative of f(x) with respect to x  dee f(x) dee x Let f(x) = x3, Find the derivative Let f(x) = x2-5x+1, Fi