breakingthisdown（breakingthisdown）.docVIP

Chapter 3 Simple RegressionRelationships In regression, we are talking about the relationships between variables. Simple regression deals with the relationship between two variables, an x (or independent variable) and a y (or dependent variable). Say, for example we wanted to study the (uninteresting) relationship between height in inches and height in feet. Graph of Feet by Inches This relationship is perfectly linear (a straight line). We can rather easily write the regression equation for this: or Recall the equation for a line: y=mx+b The intercept in this case is zero (b=0), the slope is 12 (m=12), so the regression equation can be rewritten as: y=12x+0 Slope Try to think of ?1 as the slope. That is what it is. The size of the slope is a measure of the impact of the x variable on the y. The best way to think of slope is an expression of the degree of change in y relative to the change in x. That is, for every unit change in x, there is a corresponding ?1 change in x. In the feet and inches example, every one-foot change is equal to a 12-inch change. Change in feet(x) Change in inches(y) 1 12 2 24 3 36 4 48 -1 -12 -2 -24 -3 -36 -4 -48 Assignment 14 – p105; 3.1, 3.3, 3.4, 3.5 A Slightly More Interesting Relationship The line is not quite as clear. We will be using deviation scores again and as before the sum of the deviation needs to be 0. The problem is there are a whole lot of lines that will satisfy this need. To decide on the “line of best fit” we are going to use a method called “least squares estimation.” Least Squares Estimation We could draw a lot of different lines through these data: These are the data from the graph above1.00 Var1 Var2 1 4 2 7 3 5 4 9 5 12 What are residuals? In OLS (ordinary least squares) regression, we are trying to find the line that has the smallest possible residual for each data point. A residual is the difference between the predicted value () and the