Answers to Try It Yourself questions
Slide 30:
What is the difference between a dichotomous variable and a binary variable?
Answer: A dichotomous variable takes on two values, but they can be any two values
(as well as missing, which isn't counted as a value).
A binary variable takes on ONLY the values of 0 and 1 (as well as missing).
This means that binary variables are a subset of dichotomous variables.
Why would this matter in a regression analysis?
Answer: The meaning of the intercept is still the same: The expected value
of the outcome when all predictors are held at 0. However, if the predictor is
coded 1 and 2, 0 is not in the range of the variable, so the intercept value has
little meaning.
Please see https://stats.oarc.ucla.edu/spss/webbooks/reg/chapter3/regression-with-spsschapter-3-regression-with-categorical-predictors/
Section 3.2 for more information.
Given that the interaction term is not statistically significant, should the interaction term remain in the model or be removed?
Answer: As with most questions in statistics, the answer is: It depends.
If the interaction was a test of a hypothesis, then the interaction should remain in the model.
If the interaction was part of "snooping" the data, then you may choose whether to keep it in the model or remove it.
Some researchers will remove a non-significant interaction to ease the interpretation of the lower-order terms. (page 244)
Slide 56:
What if the X variable had three levels instead of the W variable?
Write the PROCESS syntax for that model using the hsbmoderation dataset.
process y = opinion /x = detail3 /w = binaryvr /mcx = 1 /model = 1 /plot = 1.
Slide 121:
Write the SPSS GLM syntax for a model with a three-way interaction.
Answer:
glm loss with hours female effort
/design hours female effort hours*female hours*effort female*effort hours*female*effort
/print = parameter.
Why might the output not match the PROCESS output for some coefficients?
Answer: It is possible that a different reference group was used for categorical predictors.
Slide 130:
Rewrite the PROCESS syntax so that w = female and z = effort.
Answer: process y = loss /x = hours /w = female /z = effort /model = 3 /plot = 1.
Run the resulting graph code.
Do you like this graph better?
Slide 139:
What is the difference between reporting coefficients and interpreting coefficients?
Answer: Reporting the coefficients is literally just copying the coefficients
(and other values, such as SEs or CIs, test statistics, p-values, effect sizes)
into a table in your manuscript. Interpreting means that you discuss the
substantive meaning of the coefficient.
Do you need to interpret all of the coefficients in your regression output?
Answer: No.
Do you need to report all of the coefficients in your regression output, or only those that you interpret?
Answer: You need to report all of the coefficients in your output, but you can choose which cofficients to interpret.