SPSS Instructions CAUSAL ANALYSIS
General tips
Data View = shows the raw data
Variable View = provides information on the variables in the data set
The benefit of using syntax is that it provides a precise history of how the data was analysed.
For continuous variables the Frequency function is often not very useful, because many different
scores are possible. If many different values are possible, or the sample is not very big, each
respondent will have its own unique value with a frequency of one. -> Use summary statistics for these
variables, such as the mean, median, and variance
The assumptions for an ANOVA are:
Correct measurement level of the X and Y variables
Normal distribution of scores on the Y variable (for the whole data, and each group)
No outliers
Approximately equal variances in all the groups
The observations should be independent (within- and between groups)
A histogram = a bar-graph where the range of continuous scores on a variable. The frequency count of
these categories are plotted on the Y-axis (vertical) to see how often different values occur.
IV: continuous, DV: continuous -> Run Pearson’s Correlation (to see whether there is a linear
relationship)
One continuous and one categorical variable? -> Use ANOVA (to compare means)
IV: categorical or continuous, DV: continuous -> bivariate regression or multiple regression (to see
how well X explains Y)
DV dichotomous? -> use Logistic regression
A scatterplot graph = depicts every respondent as a dot on a plane, at the location of their score on the
Y and X variable. (! In some cases, when many respondents have overlapping Y and X scores, there
are less points in the graph than there are respondents since they are on top of each other. Solution ->
let the size of the dot indicate how many people are in a certain location).
Why would one prefer to use ‘recode into different variables’ instead of ‘recode into same variables’?
The Recode into Different Variables function in SPSS allows you to change the values on a variable.
In this way, if you make a mistake you still have the original variable in the data and can try again.
Predicted value = b0+ bX
Residual = Observed value - Predicted value
Multicollinearity is a term that refers to the problem whereby the independent variables share too
much variance.
the Variance Inflation Factor (VIF) in particular, test for the this potential problem by iteratively
predicting each independent variable using all other independent variables in the model. When one or
more of the X variables can be almost perfectly predicted by other variables in the model, the VIF
value becomes large for this particular predictor.
R2 change : when you have changed the sequence of which variable you add to the model; the last
model of both analyses will have similar values, because you have added the same variables to explain
the DV. The only difference however, is the R2 change, since you changed the order.
General tips
Data View = shows the raw data
Variable View = provides information on the variables in the data set
The benefit of using syntax is that it provides a precise history of how the data was analysed.
For continuous variables the Frequency function is often not very useful, because many different
scores are possible. If many different values are possible, or the sample is not very big, each
respondent will have its own unique value with a frequency of one. -> Use summary statistics for these
variables, such as the mean, median, and variance
The assumptions for an ANOVA are:
Correct measurement level of the X and Y variables
Normal distribution of scores on the Y variable (for the whole data, and each group)
No outliers
Approximately equal variances in all the groups
The observations should be independent (within- and between groups)
A histogram = a bar-graph where the range of continuous scores on a variable. The frequency count of
these categories are plotted on the Y-axis (vertical) to see how often different values occur.
IV: continuous, DV: continuous -> Run Pearson’s Correlation (to see whether there is a linear
relationship)
One continuous and one categorical variable? -> Use ANOVA (to compare means)
IV: categorical or continuous, DV: continuous -> bivariate regression or multiple regression (to see
how well X explains Y)
DV dichotomous? -> use Logistic regression
A scatterplot graph = depicts every respondent as a dot on a plane, at the location of their score on the
Y and X variable. (! In some cases, when many respondents have overlapping Y and X scores, there
are less points in the graph than there are respondents since they are on top of each other. Solution ->
let the size of the dot indicate how many people are in a certain location).
Why would one prefer to use ‘recode into different variables’ instead of ‘recode into same variables’?
The Recode into Different Variables function in SPSS allows you to change the values on a variable.
In this way, if you make a mistake you still have the original variable in the data and can try again.
Predicted value = b0+ bX
Residual = Observed value - Predicted value
Multicollinearity is a term that refers to the problem whereby the independent variables share too
much variance.
the Variance Inflation Factor (VIF) in particular, test for the this potential problem by iteratively
predicting each independent variable using all other independent variables in the model. When one or
more of the X variables can be almost perfectly predicted by other variables in the model, the VIF
value becomes large for this particular predictor.
R2 change : when you have changed the sequence of which variable you add to the model; the last
model of both analyses will have similar values, because you have added the same variables to explain
the DV. The only difference however, is the R2 change, since you changed the order.
