Step by step Manual
Multiple (hierarchical) regression analysis
1. Transform > Recode into different variables
1. Recode items in scale when needed
2. Analyze > Scale > Reliability analysis
1. Place all items of 1 scale in ‘items’
2. Statistics > Check on:
1. Scale if item deleted
2. Correlations
3. Output > reliable when values are between 0.7 and 1
4. You can only remove items when there are still 5 or more items left and when
Cronbach’s alpha increases a lot.
5. Repeat for every scale
3. Transform > Compute variable
1. Compute mean (scale) of variables.
4. Analyze > Regression > Linear
1. DV in dependent
2. Predictors in Independent(s)
1. Make multiple blocks for all the models you want to fit. You can do your hypotheses
together in 1 block and a RQ in 1 block, or all hypotheses separate, it depends on
the case (e.g. Consider whether there are some predictors that are more well-
established than others on the basis of the background information for the case &
think of how to find the model that fits your data best).
With hierarchical, SPSS automatically adds the variable in block 1 over all other
models.
3. Statistics > Check on all of them except Covariance matrix:
1. Estimates
2. Confidence intervals
3. Durbin-Watson (assumption: independence of errors)
4. Casewise diagnostics (assumption: outliers). Change number to 2.
5. Model fit
6. R squared change
7. Descriptives
8. Part and partial correlations (because you want to know if your predictors
correlate)
9. Collinearity diagnostics (assumption: multicollinearity)
4. Plots > Check on:
1. Histogram
2. Normal probability plot
(1 & 2 assumption: normality of residuals)
3. Drag *ZRESID > X (z-scores residuals), *ZPRED > Y (predicted values)
4. Produce all partial plots
(3 & 4 assumption: homoscedasticity)
5. Save > Check on:
1. Mahalanobis
2. Cook’s
Multiple (hierarchical) regression analysis
1. Transform > Recode into different variables
1. Recode items in scale when needed
2. Analyze > Scale > Reliability analysis
1. Place all items of 1 scale in ‘items’
2. Statistics > Check on:
1. Scale if item deleted
2. Correlations
3. Output > reliable when values are between 0.7 and 1
4. You can only remove items when there are still 5 or more items left and when
Cronbach’s alpha increases a lot.
5. Repeat for every scale
3. Transform > Compute variable
1. Compute mean (scale) of variables.
4. Analyze > Regression > Linear
1. DV in dependent
2. Predictors in Independent(s)
1. Make multiple blocks for all the models you want to fit. You can do your hypotheses
together in 1 block and a RQ in 1 block, or all hypotheses separate, it depends on
the case (e.g. Consider whether there are some predictors that are more well-
established than others on the basis of the background information for the case &
think of how to find the model that fits your data best).
With hierarchical, SPSS automatically adds the variable in block 1 over all other
models.
3. Statistics > Check on all of them except Covariance matrix:
1. Estimates
2. Confidence intervals
3. Durbin-Watson (assumption: independence of errors)
4. Casewise diagnostics (assumption: outliers). Change number to 2.
5. Model fit
6. R squared change
7. Descriptives
8. Part and partial correlations (because you want to know if your predictors
correlate)
9. Collinearity diagnostics (assumption: multicollinearity)
4. Plots > Check on:
1. Histogram
2. Normal probability plot
(1 & 2 assumption: normality of residuals)
3. Drag *ZRESID > X (z-scores residuals), *ZPRED > Y (predicted values)
4. Produce all partial plots
(3 & 4 assumption: homoscedasticity)
5. Save > Check on:
1. Mahalanobis
2. Cook’s
