Summary Multivariate data analyse (MVDA) SPSS stappenplan

WG1 – Linear Regression
Pearson correlation
Analyze  Correlate  Bivariate



Linear regression analysis
Analyze  Regression  Linear  Dependent: afhankelijke variabele  Independents: alle
onafhankelijke variabelen
 Statistics: Part & partial correlations (VAF per variabele)
 Statistics: Collinearity diagnostics (multicollineariteit)
o VIF < 10 (geen multicollineariteit)
o Tolerance > .10 (geen multicollineariteit)
 Save: Cook’s distance & Leverage values (outliers)
o Cook’s distance < 1 (geen outliers)
o Leverage value < 3(k+1)/N (geen outliers)
 Plots: scatterplot van standardized predicted values (Y) vs. standardized residuals (X)
o Vraag normal probability plot


Hierarchical regression analysis
Analyze  Regression  Linear  Dependent: afhankelijke variabele  Independents: variabelen
op bepaalde volgorde in de losse blokken
 Statistics: R2change
 Zie Linear Regression Analysis


Remove outliers
(Ask for Cook’s distances and Leverage when performing linear regression)  Data  Select cases 
If condition is satisfied  If  Cook’s Distance < 1



Mogelijke vragen?
 What predictor explains the most unique variance?  part correlations
 How much variance of [Y] is explained by [X]?  (part correlations)2
 Assumpties checken
o Non-linearity: scatterplot
o Heteroscedasticity: scatterplot
o Non-normality of the residuals: normal probability plot
o Multicollineariteit: collinearity diagnostics
 How much variance of [Y] is explained by the predictors?  Model summary: R2

, WG2 – ANOVA
Crosstabs (group sizes)
Analyze  Descriptive statistics  Crosstabs



ANOVA
Analyze  General linear models  Univariate
 Plots: variabele met meeste groepen in horizontal axis, andere groep in separate lines
(interaction effect interpreteren) (“profile plot”)
 EM means: estimated marginal means (losse variabelen & interactie effect)
 Save: standardized residuals (+histogram voor assumpties (zie hieronder))
 Options: descriptive statistics & homogeneity tests
 Post hoc: Tukey



Distribution standardized residuals (voor assumpties)
ANOVA  Save: standardized residuals  OK  Graphs  Histogram of standardized residuals



Mogelijke vragen?
 Assumpties checken
o Normaliteit: histogram per groep (distribution standardized residuals) of
Kolomogorov-Smirnov
o Equality group variances (homogeneity): Levene’s test > .05 goed
 Is there a significant relationship between Y and X 1 and/or X2?  Tests of Between-Subjects
Effects: Corrected model
 Can the null hypothesis of equal group variances be rejected?  Levene’s test
 Calculate the eta squared measure  formuleblad



WG3 – ANCOVA
Drie stappen:
1. ANOVA om te checken of x effectief is  als significant:
2. ANCOVA met interactie effect om parallelisme assumptie te checken
a. ANCOVA  Univariate  Model  Build model
i. Als significant: niet voldaan aan assumptie parallelle regressielijnen
ii. Als niet-significant: stap 3
3. ANCOVA zonder interactie



“Determine the correlation between Covariate and Dependent Variable”
ANOVA  Dependent: Covariate  Independent: Dependent variable  Output: Corrected model