Chapter 11: The R.M.S. Error for Regression Errors: A has a large - - PowerPoint PPT Presentation

Chapter 11: The R.M.S. Error for Regression

Errors: A has a large positive error B has a large positive error C has a negative error D has a negative error E has a positive error

The r.m.s. error is the r.m.s. size of the errors. The r.m.s. error measures how good a prediction

• is. It says how large the errors are likely to be.

To calculate the r.m.s. error use the following shortcut:

r.m.s. error = √(1 – r2) (SDY)

Example 1: For the men aged 18-24 in the HANES sample, the relationship between height and systolic blood pressure can be summarized as follows: Average height ≈ 70”, SD ≈ 3” Average b.p. ≈ 124mm, SD ≈ 14mm r = -0.2

a)

Estimate the average blood pressure of men who were 6 feet tall.

b)

Find the r.m.s. error of the prediction

If the scatter diagram is football-shaped, the r.m.s. error is like an SD for the regression line.

68% of the dots fall between the line ± 1 r.m.s. error 95% of the dots fall between the line ± 2 r.m.s. errors

One r.m.s. error up and down

68%

Two r.m.s. errors up and down

95%

If the scatter diagram is football-shaped, the r.m.s. error says how far a typical point is above or below the regression line. It gives us a give-or-take number for our estimates.

Example 2: Midterm: ave = 65 SD = 16 r = 0.7 Final: ave = 60 SD = 10 Estimate the final exam score for someone who got 81 on the midterm and put a give-or-take number on your estimate.

If the scatter diagram is football-shaped, the r.m.s. error can be used like an SD for the regression line. Approximately 95% of the points will be between regression estimate – 2(r.m.s. error) and regression estimate + 2(r.m.s. error)

Example 3: Midterm: ave = 65 SD = 16 r = 0.7 Final: ave = 60 SD = 10 Estimate the final exam score for someone who got 81 on the midterm. Would you be surprised to hear that the student scored 70? How about 77? 60?

Residuals

The residual says how far the point is above or below the line. To see if the scatter diagram is football-shaped, we plot the residuals

Residual plots make it easier to see if the scatter diagram is football-shaped. Is this one football- shaped?

If the residual plot has a pattern, the regression might not be appropriate.

A football-shaped scatter diagram is said to be “________________”. A scatter diagram that has more variability on one side is said to be “_______________”. Which is this?

The r.m.s. error is only appropriate for ____________ scatter diagrams. If you don’t know what the scatter diagram looks like, it is dangerous to do the

• regression. In this case, you have to assume

that it is football-shaped and if this assumption is incorrect, your answers may not be accurate.