# Minitab Help 10: Model Building

Minitab Help 10: Model Building##
Minitab^{®}

## Martians (underspecified model)

- Perform a linear regression analysis of weight vs height + water. Click "Storage" and select "Fits" before clicking "OK."
- Select Calc > Calculator, type "FITS0" in the box labeled "Store results in variable," type "if(water=0, FITS)" in the box labeled "Expression," and click "OK." Repeat to create "FITS10" as "if(water=10,FITS)" and "FITS20" as "if(water=20,FITS)."
- Perform a linear regression analysis of weight vs height (an underspecified model). Click "Storage" and select "Fits" before clicking "OK." The resulting variable should be called "FITS_1."
- Create a basic scatterplot but select "With Groups" instead of "Simple." Plot "weight" vs "height" with "water" as the "Categorical variable for grouping."
- To add parallel regression lines representing the different levels of water to the scatterplot, select the scatterplot, select Editor > Add > Calculated Line, and select "FITS0" for the "Y column" and "height" for the "X column." Repeat to add the "FITS10" and "FITS20" lines.
- To add a regression line representing the underspecified model to the scatterplot, select the scatterplot, select Editor > Add > Calculated Line, and select "FITS_1" for the "Y column" and "height" for the "X column."

## Cement hardening (variable selection using stepwise regression)

- Create a simple matrix of scatterplots of the data.
- Conduct stepwise regression for y vs \(x_1\) + \(x_2\) + \(x_3\) + \(x_4\).

## IQ and body size (variable selection using stepwise regression)

- Create a simple matrix of scatterplots of the data.
- Conduct stepwise regression for PIQ vs Brain + Height + Weight.

## Blood pressure (variable selection using stepwise regression)

- Create a simple matrix of scatterplots of the data.
- Conduct stepwise regression for BP vs Age + Weight + BSA + Dur + Pulse + Stress.

## Cement hardening (variable selection using best subsets regression)

- Conduct best subsets regression for y vs \(x_1\) + \(x_2\) + \(x_3\) + \(x_4\).
- Perform a linear regression analysis of all four predictors (assumed unbiased) and just two predictors to retrieve the information needed to calculate \(C_p\) for the model with just two predictors by hand.
- Perform a linear regression analysis of model with \(x_1\), \(x_2\), and \(x_4\) and note the variance inflation factors for \(x_2\) and \(x_4\) are very high.
- Perform a linear regression analysis of the model with \(x_1\), \(x_2\), and \(x_3\) and note the variance inflation factors are acceptable.
- Perform a linear regression analysis of model with \(x_1\) and \(x_2\) and note the variance inflation factors are acceptable and adjusted \(R^{2}\) is high
- Create residual plots and Conduct regression error normality tests.

## IQ and body size (variable selection using best subsets regression)

- Conduct best subsets regression for PIQ vs Brain + Height + Weight.
- Perform a linear regression analysis of the model with Brain and Height and note the variance inflation factors are acceptable and adjusted \(R^{2}\) is as good as it gets with this dataset.
- Create residual plots and Conduct regression error normality tests.

## Blood pressure (variable selection using best subsets regression)

- Conduct best subsets regression for BP vs Age + Weight + BSA + Dur + Pulse + Stress.
- Perform a linear regression analysis of the model with Age and Weight and note the variance inflation factors are acceptable and adjusted \(R^{2}\) can't get much better.
- Create residual plots and Conduct regression error normality tests.

## Peruvian blood pressure (variable selection using best subsets regression)

- Select Calc > Calculator to create fraclife variable equal to Years/Age.
- Conduct best subsets regression for Systol vs Age + Years + fraclife + Weight + Height + Chin + Forearm + Pulse.
- Perform a linear regression analysis of the best 5-predictor and 4-predictor models.
- Calculate AIC and BIC by hand.

## Measurements of college students (variable selection using stepwise regression)

- Conduct stepwise regression for Height vs LeftArm + LeftFoot + LeftHand + HeadCirc + nose + Gender.
- Change the "Method" to "Backward elimination" or "Forward selection."