This page includes Step-by-Step instructions to use Minitab to calculate Best Subsets Regression  for the Water/Temperature example.  

Note: If you have not used Minitab before, you may want to try the Introduction to Minitab tutorial.

Example: Assume that during a three-hour period spent outside, a person recorded the temperature, the time spent mowing the lawn, weather there was sun or not (0 or 1) and their water consumption.  Using our imagination we can come up with other possible predictor variables that may make our model more accurate, like the temperature squared and the temperature times the mowing time.  Use Minitab to calculate the best model to use for predicting the amount of water consumption.

Temperature

Mowing Time

Water

Only sun

Temp2

 Temp*Mowing Time
75 1.85 16 1 5625 138.75
83 1.25 20 0 6889 103.75
85 1.5 25 0 7225 127.5
85 1.75 27 1 7225 148.75
92 1.15 32 0 8464 105.80
97 1.75 48 1 9409 169.75
99 1.6 48 0 9801 158.40

 

Preliminaries:  Entering Data

  1. Enter “temp” in the title row of column C1.
  2. Enter "mow time" in the title row of column C2.
  3. Enter “water” in title row of column C3.
  4. Enter "only sun" in the title row of column C4.
  5. Enter "temp^2" in the title row of column C5.
  6. Enter "temp*mow" in the title row of column C6.
  7. Enter the data in the appropriate columns.  Do not change the order of the items.  Numerical values should be right justified. If they are not, then click here for more information.

  1. Save the data. Select FILE > SAVE AS. Call the data set BestSubsetsData. (You may use this data again in future tutorials.)

Procedure:  Best Subsets Regression 

  1. From the main menu bar select Stat à Regression à Best Subsets.
  2. Select the response variable to be water.
  3. Select the Free Predictors.  You may select up to 20 free predictors.  If there is a predictor that must be included in every model then put that variable in the box labeled 'Predictors in all models'.  However, our example does not have any predictors that we feel must be included. 

4.Click on Options.  Here you can change the options of Best Subsets Regression.  The minimum field is the minimum number of free predictors that must be included in each model.  The maximum field is the maximum number of free predictors that may be included in each model.  You can also change how many models of each size get displayed by changing the value of the box labeled “Models of each size to print”.  By unchecking the “Fit Intercept” box you can exclude the y-intercept (B0) from the regression model.  For our example we will use the default values for all options.

5.  Click OK.

6.Click OK.

    

  Solution Interpretation  

Regression Tutorial Menu  Dictionary

STATS @ MTSU