Flashcards in Chapter 7: Linear Regression Deck (11):
Regression definition go!
- uses relationship between variables for prediction
- two variables are involved....the predictor and the criterion
- variables are not equal
Least square criterion?
Y prime= predicted y value
Y= observed Y value
Y-Y prime = prediction error for each data point
Least Squares regression line?
- line that minimizes total prediction error
- all vert distances should be in their absolute min
-only one line....
Least squares regression line formula?
- Y'= byX+ay
Y'= predicted value
ay= y-axis intercept for min error in y
X= any value of x
How to make the regression line?
solve Y' for the lowest and highest x values.
- plot the value and connect the two points.
calc y int and connect line to the int
What is the best predictor of Y???
the regression line
- the weaker the relationshup the more the scores vary from the regression line
the stronger the regression line =the closer
Size wise what do you want the standard error of the esitmate to be?
- want it to be small
Characteristics of the standard error of the estimate?
- treat it like standard deviation for predictions of percents..
+1/-1 = 68%
+/- 2= 95%
+/- 3= 99%
- basing std error on the spread
- best case