回归与相关-Southeast.PPT

* Regression Equation a is intercept，the value of y when X=0； b, slope, 。 b, regression coefficient, the average units that y change when x change by 1 units. * Steps of regression analysis 1.Scatter plot for linear trend 2.Estimate slope and intercept。 3.Draw regression line 4.Significant test for b * Least Square Estimation (LSE) principle：assure the sum of squared differences between Y’s and their estimates( residuals) based on regression line , that is least square estimation of regression parameters. we want to find a and b that the sum of squares about the line is a minimum.we there fore want to minimize . this will have a minimum when the partial differentials with respect to a and b are both zero. if all points lay aling a line and there was no random variation,it would be easy to draw a line on the scatter diagram.but it isnt the case.there are many possible value of a and b,we need a criterion for choosign the best line. * Least Square Estimation Y adult height（inch） X height of 2 years old 30 32 34 36 38 40 63 65 67 69 71 this figure shows the deviationof a point from the line, the distance from the point to the line in the y direction. if the deviation from it are small the line willfit the data well,and will fit badly if they are large. this deviation represent the part of y not explained by x. one solution to the problem of finding the best line is to choose that which leaves the ninimum amount of the variability of y unexplained, by making the variance of error a ninimum. this will be achived by making the sum of squres of the diviations about the line a minimum. this is called the method of least squares,the line found is the least squares line. * Data no 2 years (inch) X adult (inch) Y x2 y2 xy 1 39 71 2 30 63 3 32 63 4 34 67 5 35 68 6 36 68 7 36 70 8 30 64 * Relationship between 2 heights variables * Draw regression line Choose 2 x points that are not too close and easy to measure, calculate the estimates based on the equation. For