Lecture 9:
- Which correlation measures do you know if your variables are interval? And which ones if
one or both variables are ordinal?
For interval variables the correlation measures Pearson’s r, Spearman’s rho and Kendall’s tau
suffice. When there are ordinal variables involved, only Spearman’s rho and Kendall’s tau can be
used.
- Does it make a difference for the measures eta, r, rS, tau if you switch the role X and Y?
Switching the role of X and Y makes no difference for r, rs and eta. It does make a difference for
the measure tau.
- What does it mean if you find an rs of -0.3?
An rs of -0.3 means there is a negative correlation, due to the minus sign. A 1 unit increase in X
associates with a 0.3 decrease in Y. It also means there is medium linear association, since rs2 =
0.09.
- What does the partical correlation coefficient rXY.W reflect?
The partial correlation coefficient rx,y,w reflects the correlation between X and Y controlled for by
the variable W. This can have all sorts of effects on the relation between X and Y.
Lecture 10:
- How do you interpret the coefficient in a simple regression?
The coefficient in a simple regression is interpreted through the equation Y = β0 + β1X1 + error for
a linear regression. Also the regression coefficient is represented by (𝑥, 𝑦). It is explained by the
variation levels.
- How do you interpret a coefficient in a multiple regression?
In a multiple regression we use Y = (β0 + β1𝑋1 +β2𝑋2 + β3𝑋𝑥 +...) + error and the least squares Σ
e2-method. We can interpret the coefficient via the partial and part correlation
- How can you test H0 for variables?
Via the exceedance probability, the confidence interval and the critical value. In a multiple
regression you can look at the F-value.
- How do you determine which variable has the largest partial impact?
By determining which variable has the largest partial correlation coefficient and therefore has
the largest contribution. This is determined by which one has the largest standardized
coefficient.
- What do the correlations Zero-order, Partial en Part tell you?
Zero-order is a correlation between 2 variables without controlling for the influence of any other
variable. Partial indicates the correlation between X and Y, after eliminating the influence of
other predictors on X and Y. Part relates to the correlation between X and Y, after eliminating
the influence of other predictors on just X.
- How can you evaluate the model?
The model can be evaluated by R2, R2 change and adjusted R2
- How can you include a nominal x-variable with five categories in a multiple regression?
You can include a nominal x-variable with 5 categories in a multiple regression by making
dummy variables from it.
- How can you test for interaction effects in a multiple regression?
You can test for interaction effects in a multiple regression by including the interaction term in
the model. Then you look at the F-values in the ANOVA-table, the R2 and the R2 change. Hereby,
you evaluate the model and the individual predictors.
- Which correlation measures do you know if your variables are interval? And which ones if
one or both variables are ordinal?
For interval variables the correlation measures Pearson’s r, Spearman’s rho and Kendall’s tau
suffice. When there are ordinal variables involved, only Spearman’s rho and Kendall’s tau can be
used.
- Does it make a difference for the measures eta, r, rS, tau if you switch the role X and Y?
Switching the role of X and Y makes no difference for r, rs and eta. It does make a difference for
the measure tau.
- What does it mean if you find an rs of -0.3?
An rs of -0.3 means there is a negative correlation, due to the minus sign. A 1 unit increase in X
associates with a 0.3 decrease in Y. It also means there is medium linear association, since rs2 =
0.09.
- What does the partical correlation coefficient rXY.W reflect?
The partial correlation coefficient rx,y,w reflects the correlation between X and Y controlled for by
the variable W. This can have all sorts of effects on the relation between X and Y.
Lecture 10:
- How do you interpret the coefficient in a simple regression?
The coefficient in a simple regression is interpreted through the equation Y = β0 + β1X1 + error for
a linear regression. Also the regression coefficient is represented by (𝑥, 𝑦). It is explained by the
variation levels.
- How do you interpret a coefficient in a multiple regression?
In a multiple regression we use Y = (β0 + β1𝑋1 +β2𝑋2 + β3𝑋𝑥 +...) + error and the least squares Σ
e2-method. We can interpret the coefficient via the partial and part correlation
- How can you test H0 for variables?
Via the exceedance probability, the confidence interval and the critical value. In a multiple
regression you can look at the F-value.
- How do you determine which variable has the largest partial impact?
By determining which variable has the largest partial correlation coefficient and therefore has
the largest contribution. This is determined by which one has the largest standardized
coefficient.
- What do the correlations Zero-order, Partial en Part tell you?
Zero-order is a correlation between 2 variables without controlling for the influence of any other
variable. Partial indicates the correlation between X and Y, after eliminating the influence of
other predictors on X and Y. Part relates to the correlation between X and Y, after eliminating
the influence of other predictors on just X.
- How can you evaluate the model?
The model can be evaluated by R2, R2 change and adjusted R2
- How can you include a nominal x-variable with five categories in a multiple regression?
You can include a nominal x-variable with 5 categories in a multiple regression by making
dummy variables from it.
- How can you test for interaction effects in a multiple regression?
You can test for interaction effects in a multiple regression by including the interaction term in
the model. Then you look at the F-values in the ANOVA-table, the R2 and the R2 change. Hereby,
you evaluate the model and the individual predictors.
