Chapter 29 Week 6 - Reading questions
Questions:
- What is according to Baron and Kenny the nature of a moderation effect? (and can you give an example from an article or theory you had in a previous course?
- List the four methods hat according to Baron and Kenny are available to study interaction effects in SPSS given different levels of measurement (categorical or continuous) for the independent and moderating variable?
- On page 1177, Baron Kenny mention the most common approach to deal with unreliability in the mediating variable is to use multiple indicators. Thinking back to the readings and lectures about factor analysis, can you shortly explain why this is a good method?
- The framework that Baron and Kenny present on page 1179 is one way to study complicated models (mediating moderation, or moderating mediation). They do not discuss alternatives to this framework, but (surprise!) Structural Equation Modeling is one of them. What would you think be the main advantage of SEM over the framework of Baropn and Kenny.
- How can you determine whether a variable is a mediator or moderator?
Reference
Notes before reading:
- To all you non-counseling psychologists: this article is very general, don’t worry
- this article is quite general, and you might recognize a lot of things we have discussed. This article provides an overview of things we discussed, but also adds an important aspect: the combination of factor model and path models into one Structural Equation Model
- skip the part on GFI (p. 741) – this index has been shown to be dependent on sample size after all and is old-fashioned (in a bad way, in statistics we don’t do vintage).
- skip the part on missing data – there is nothing wrong with this section, but missing data is a very difficult topic, and there is a lot more to say about this.
Take the course ”Conducting a Survey” if you want to know more on missing data!
Questions:
- The authors (Weston and Gore) state three similarities and two big differences between SEM and other multivariate statistical techniques (ANCOVA, regression). What are they?
- The second difference that Weston and Gore mention is presented as being a disadvantage, while some would think of it as an advantage. Tests for the model fit can indicate whether your theory can be confirmed or rejected based on your data! Along with this, the authors miss another big difference between SEM and other general linear model (ANOVA, regression). Any idea what this is?
- Structural Equation Models are called hybrid models when they contain a measurement and structural model. Can you explain what is meant by the terms “measurement” and “structural” model?
- On p. 726-726, the authors introduce the term item parcels. Look up (on the internet) what is meant with an item parcel. Can you think of an example of item parceling that you used during on of the practicals of this course? Revisit this practical, and try to thank about the disadvantages of item parceling in SEM.
- The authors identify 6 steps in doing SEM-analyses. What are they?
- Go back to the unconstrained multigroup path model estimated in the class exercise of practical 6 (question 8. last week). Try to work out for yourself how many degrees of freedom there are based on pages 732-733 of the article.
- On page 745 the authors mention in the 6th line from the bottom of the page, that it is a good idea to test the model using cross-validation. Look up what cross-validation in SEM is about, and explain when it is a good idea to do a cross-validation check
- The model that the authors use as an example does not fit the data very well (see figure 4 on page 740) and the discussion of the authors of their model. Without having the data, nor a good theories on the constructs used in this paper, can you think of a possible reason why this model might not fit? (there is no one right answer here). Hint: have a look at the results that are presented and those not presented in the paper