2: Moderating effects

Question 1

How do you reject the null hypothesis?

Answer 1

• You cannot know if it is true or false, there is always a risk of error.
• Look at the p-value, which shows the significance of variables. Shows how likely you are to be wrong if you reject the null hypothesis (never exactly 0).

Question 2

How is moderating effects characterised?

Answer 2

As an interaction

Question 3

If the data is expanding in a graph, what can it be a sign of?

Answer 3

Heteroscedasticity (st.dev. increases with the x-variable), which can imply a there is a moderating effect

Question 4

What is another word for beta?

Answer 4

The coefficient

Question 5

What do we assume about betas in the original OLS model?

Answer 5

That they are constant: the same no matter the values of the Xs

Question 6

What do we claim about a beta when considering moderation?

Answer 6

That the beta of the independent variable is a function of the moderating variable, hence not constant

Question 7

What does a significant p-value imply?

Answer 7

That the coefficient is significantly different from 0 (Null hypothesis: beta=0)

Question 8

What do you have to do if you have multiple continuous or categorical variables in an interaction?

Answer 8

• Standardize (z-score) your continuous variables, i.e. you must center them to 0. Reduces multicollinearity issues.
• Main effects cannot be interpreted alone! If an interaction is present, you must consider the interaction in your interpretation.
• Can also run interactions with multiple categorical or continuous variables together.

Question 9

What is factorial design and when is it used?

Answer 9

Multiple interactions together. Used in experimental design when multiple factors work together in a different way than each of the factors would work individually.

Question 10

What does a higher adjusted R2 mean?

Answer 10

Our independent variables explain more of the variance in our dependent variable. The model has a better fit: still the same data but the model explains it better

Question 11

What is moderation?

Answer 11

• The relationship between the IV and the DV is contingent upon another variable (the value of the moderating variable).
• Technically, it is an interactive effect between the IV and the moderating variable that is regressed on the DV

Question 12

What are the five steps in estimating the moderating effect?

Answer 12

1. An interaction term
2. Z-score when the moderator is continuous.
3. Mean-center (or z-standardize) both IV and moderator if continuous.
4. Create the interaction term IV*M
5. Regress the DV on X, M and X*M

Question 13

What are three different types of moderation?

Answer 13

1. Exponential moderation (likely to have heteroscedasticity issues).
2. Antagonistic moderation. ß2X2 when X2 is a dummy variable, gives two interceptions on the Y-axis.
   Intercept if X2 = 0 → alpha
   Intercept if X2 = 1 → alpha + ß2
   → likely that ß2 is significant since lines are not crossing the y-axis in the same point.
   → likely that original slope is not significant.
3. Crossing moderation. Lines look like a cross.
   → likely that original slope is not significant.

Question 14

What is econometrics?

Answer 14

• Understand the world by confronting theory to data
• Testing hypothesis inferred from theory
• Often interested in cause and mechanism