# Chapter 10: Multifactorial analyses using SPSS

###
Figure 10.1 *Output for one-way analysis of variance*

###
Box 10.3 *Presenting the results for one-way analysis of variance*

### Figure 10.1 and Box 10.3 Instruction

Click here to show the instructions as text.

- Open the file.
- From the
**File**menu choose**Open > Data...**. - Find the file
*smokingbwt.sav*and open it. - Tell SPSS to do an ANOVA.
- From the
**Analyze**menu choose**Compare Means > One-Way ANOVA...**. - Drag
*birthwt*to the**Dependent List:**box. - Drag
*smokegroup*to the**Factor:**box. - Click
**Options...**. - In the
**Statistics**seciton check on the**Descriptive**checkbox. - Click
**Continue**. - Click
**OK**.

###
Figure 10.2 *Output for multiple regression*

###
Box 10.5 *Presenting the results of multiple regression*

### Figure 10.2 and Box 10.5 Instruction

Click here to show the instructions as text.

- Open the file.
- From the
**File**menu choose**Open > Data...**. - Find the file
*perf.sav*and open it. - Tell SPSS to do a multiple regression.
- From the
**Analyze**menu choose**Regression > Linear...**. - If a window appears asking you to set measurement level, click
**OK**. - Drag
*perf*to the**Dependent:**box. - Drag
*age*,*gas*,*smokers*,*wheeze*to the**Independent(s):**box. - Click
**Statistics...**. - In the
**Regression Coefficients**section check the**Confidence intervals**checkbox. - Uncheck the
**Model fit**checkbox. - Click
**Continue**. - Click
**OK**. - Tell SPSS to do a multiple regression with smokers dropped.
- From the
**Analyze**menu choose**Regression > Linear...**. - If a window appears asking you to set measurement level, click
**OK**. - Drag
*perf*to the**Dependent:**box. - Drag
*age*,*gas*and*wheeze*to the**Independent(s):**box. - Click
**Statistics...**. - In the
**Regression Coefficients**section check the**Confidence intervals**checkbox. - Uncheck the
**Model fit**checkbox. - Click
**Continue**. - Click
**OK**. - Tell SPSS to do a multiple regression with smokers and wheeze dropped.
- From the
**Analyze**menu choose**Regression > Linear...**. - If a window appears asking you to set measurement level, click
**OK**. - Drag
*perf*to the**Dependent:**box. - Drag
*age*and*gas*to the**Independent(s):**box. - Click
**Statistics...**. - In the
**Regression Coefficients**section check the**Confidence intervals**checkbox. - Uncheck the
**Model fit**checkbox. - Click
**Continue**. - Click
**OK**.

###
Figure 10.3 *Output from multiple regression with categorical predictor variables*

###
Box 10.8 *Presentation of multiple regression with categorical predictor variable*

### Figure 10.3 and Box 10.8 Instruction

Click here to show the instructions as text.SPSS does not automatically code categorical predictors into a set of (dummy) predictor variables when it does regression. The instructions below explain how to make them.

- Open the file.
- From the
**File**menu choose**Open > Data...**. - Find the file
*smokingbwt.sav*and open it. - First check to see if there is missing predictor data.
- From the
**Analyze**menu choose**Descritpive Statistics > Frequencies...**. - Drag
*smokegroup*and*alcgroup*to the**Variable(s)**box. - Click
**OK**. - Exclude records that have missing values..
- From the
**Data**menu choose**Select cases...**. - In the
**Select**section, choose the**If condition is satisfied**radio button. - Click
**If...**. - In the box at the top of the
**Select cases: If**type*not (sysmis(cigsgp) | sysmis(alcgroup))*. That is, select the records where the data is not missing alcohol or smoking group data. - Click
**OK**. - Make a 0 or 1 coded dummy variable indicating who is a heavy smoker. Note that you don't need to make a dummy variable indicator of who is not a heavy smoker because you know it using this one variable.
- From the
**Transform**menu choose**Recode into different variables...**. - Drag
*alcgroup*to the**String Variable -> Output Varible:**box. - Click
**Old and New Values...**. - Type
*2*in the**Old Value Value:**box. - Type
*0*in the**New Value Value:**box. - Click
**Add**. - Type
*1*in the**Old Value Value:**box. - Type
*1*in the**New Value Value:**box. - Click
**Add**. - Click
**Continue**. - Type
*LightSmoker*in the**Output Variable Name:**box. - Type
*A light smoker*in the**Output Label Name:**box. - Click
**Change**. - Click
**OK**. - Predict adjusted baby weight with smoking group and alcohol (using a linear model / ANOVA).
- From the
**Analyze**menu choose**General Linear Model > Univariate...**. - Drag
*adjbw*to the**Dependent Variable:**box. - Drag
*lightSmoker*and*alcgroup*to the**Fixed Factor(s):**box. - Click
**OK**. - Make a set of 0 or 1 coded dummy variables indicating the amount of alcohol consumed by drinks. Note that you don't need to make a dummy variable indicator of who does not drink any alcohol because that person would be coded with zeros on all the drinking (amount) variables.
- From the
**Transform**menu choose**Recode into different variables...**. - Drag
*alcgroup*to the**String Variable -> Output Varible:**box. - Click
**Old and New Values...**. - Type
*1*in the**Old Value Value:**box. - Type
*1*in the**New Value Value:**box. - Click
**Add**. - Click the
**All other values**radio button. - Type
*0*in the**New Value Value:**box. - Click
**Add**. - Click
**Continue**. - Type
*drink1_19*in the**Output Variable Name:**box. - Type
*Drinks 1 to 19 g per week*in the**Output Label Name:**box. - Click
**Change**. - Click
**OK**. - From the
**Transform**menu choose**Recode into different variables...**. - Drag
*alcgroup*to the**String Variable -> Output Varible:**box. - Click
**Old and New Values...**. - Type
*2*in the**Old Value Value:**box. - Type
*1*in the**New Value Value:**box. - Click
**Add**. - Click the
**All other values**radio button. - Type
*0*in the**New Value Value:**box. - Click
**Add**. - Click
**Continue**. - Type
*drink20_49*in the**Output Variable Name:**box. - Type
*Drinks 20 to 49 g per week*in the**Output Label Name:**box. - Click
**Change**. - Click
**OK**. - From the
**Transform**menu choose**Recode into different variables...**. - Drag
*alcgroup*to the**String Variable -> Output Varible:**box. - Click
**Old and New Values...**. - Type
*3*in the**Old Value Value:**box. - Type
*1*in the**New Value Value:**box. - Click
**Add**. - Click the
**All other values**radio button. - Type
*0*in the**New Value Value:**box. - Click
**Add**. - Click
**Continue**. - Type
*alcgroup*in the**Output Variable Name:**box. - Type
*Drinks 50 to 99 g per week*in the**Output Label Name:**box. - Click
**Change**. - Click
**OK**. - From the
**Transform**menu choose**Recode into different variables...**. - Drag
*alcgroup*to the**String Variable -> Output Varible:**box. - Click
**Old and New Values...**. - Type
*4*in the**Old Value Value:**box. - Type
*1*in the**New Value Value:**box. - Click
**Add**. - Click the
**All other values**radio button. - Type
*0*in the**New Value Value:**box. - Click
**Add**. - Click
**Continue**. - Type
*drink100*in the**Output Variable Name:**box. - Type
*Drinks 100+ g per week*in the**Output Label Name:**box. - Click
**Change**. - Click
**OK**. - Tell SPSS to do a multiple regression with the dummy coded categorical variables.
- From the
**Analyze**menu choose**Regression > Linear...**. - If a window appears asking you to set measurement level, click
**OK**. - Drag
*adjbw*to the**Dependent:**box. - Drag
*HeavySmoker*,*drink1_19*,*drink20_49*,*drink50_99*, and*drink100*to the**Independent(s):**box. - Click
**Statistics...**. - In the
**Regression Coefficients**section check the**Confidence intervals**checkbox. - Uncheck the
**Model fit**checkbox. - Click
**Continue**. - Click
**OK**.

###
Figure 10.5 *Output for logistic regression*

###
Box 10.12 *Presenting the results for logistic regression*

### Figure 10.5 and Box 10.2 Instruction

Click here to show the instructions as text.

- Open the file.
- From the
**File**menu choose**Open > Data...**. - Find the file
*bpd.sav*and open it. - Tell SPSS to do logistic regression to predict use of drugs with 4 variables.
- From the
**Analyze**menu choose**Regression > Binary Logistic...**. - Drag
*chest*to the**Dependent:**box. - Drag
*sex*,*multiple*,*o2dep36w*,*o2depdis*to the**Dependent:**box. - Click
**Categorical...**. - Drag
*sex*and*multiple*to the**Categorical Covariates:**box. - Set
**Contrast**to be**Simple** - Push the
**First**radio button. - Click
**Change**and make sure both variables should be listed with**(indicator(first))**. Repeat the steps above on each variable if needed. - Click
**Continue**. - Click
**Options...**. - In the
**Statistics and Plots**section check the**CI for exp(B)**checkbox. - Click
**Continue**. - Click
**OK**. - Tell SPSS to do logistic regression to predict use of drugs with 3 variables.
- From the
**Analyze**menu choose**Regression > Binary Logistic...**. - Drag
*chest*to the**Dependent:**box. - Drag
*sex*,*multiple*and*o2dep36w*to the**Dependent:**box. - Click
**Categorical...**. - Drag
*sex*and*multiple*to the**Categorical Covariates:**box. - Set
**Contrast**to be**Simple**. - Push the
**First**radio button. - Click
**Change**and make sure both variables should be listed with**(indicator(first))**. Repeat the steps above on each variable if needed. - Click
**Continue**. - Click
**Options...**. - In the
**Statistics and Plots**section check on the**CI for exp(B)**checkbox. - Click
**Continue**. - Click
**OK**.

###
Figure 10.6 *Output for logistic regression with categorical predictor variables*

###
Box 10.14 *Presenting logistic regression with categorical predictor variables*

### Figure 10.6 Instruction

Click here to show the instructions as text.

- Open the file.
- From the
**File**menu choose**Open > Data...**. - Find the file
*smokingbwt.sav*and open it. - Tell SPSS to do logistic regression to predict use of drugs with 2 categorical variables.
- From the
**Analyze**menu choose**Regression > Binary Logistic...**. - Drag
*nausea*to the**Dependent:**box. - Drag
*smoker1*and*socialclass*to the**Covariates:**box. - Click
**Categorical...**. - Drag
*smoker1*and*socialclass*to the**Categorical Covariates:**box. - Set
**Contrast**to be**Simple** - Push the
**First**radio button. - Click
**Change**and make sure both variables should be listed with**(indicator(first))**. Repeat the steps above on each variable if needed. - Click
**Continue**. - Click
**Options...**. - In the
**Statistics and Plots**section check on the**CI for exp(B)**checkbox. - Click
**Continue**. - Click
**OK**.

###
Box 10.15 *Presenting unadjusted and adjusted logistic regression*

### Box 10.15 Instruction

SPSS can not easily make this output.