Exercise 13 No published assessment instruments exist to examine attitudes toward harm reduction (HR)

Exercise 13

No published assessment instruments exist to examine attitudes toward harm reduction (HR) and abstinence only (AO) alcohol intervention techniques. Therefore, the authors generated 46 items to assess participants’ attitudes toward the effectiveness of techniques that primarily are associated with either HR (e.g., "when you drink, try and space your drinks out") or AO (e.g., "completely avoid places you used to drink") interventions. Items representing the intervention strategies were rated by participants for perceived effectiveness on a five point scale from 1 ( not at all likely to be effective ) to 5 ( very likely to be effective ). Participants were provided the following directions: "There are many ways to manage problematic alcohol use. Imagine that you have an alcohol problem. Below are different types of things that you could do to try to change your problem. Rate how effective you think each strategy would be if you had an alcohol problem."

Using the data file provided: STATS Exercise 13

Conduct an Exploratory Factor Analysis (EFA) on the scale items using the steps provided below. Please note that the aim of the scale was to create two factors, which you might consider in addition to the scree plot when deciding on the number of factors to ultimately retain. Provide a write-up of your results with a table (see provided example based on the Revised Firewater Myth Scale) and your output. In your write-up, be sure to note which items you are recommending not be retained on the newly developed Treatment Attitudes Scale and why.

In addition answer the following questions:

How do we interpret (what is) a communality?

How do we interpret the factor loadings?

What is common variance vs. unique variance in EFA? Which one do we hope is maximized?

What is presumed to be causing the measured variables?

Outline of steps in Factor Analysis and User Decisions

Step 1 : User decides on "method of extraction". Two methods of extraction of loadings are described book PC and EFA (specifically via PAF)
- For the most part, we would use EFA (as opposed to PC) and PAF as the method of extraction
  - Maximum Likelihood is another recommended method of extraction
Step 2 : User must decide on criterion for the number of Factors to retain.
- There is a default criterion in SPSS (only retain components or factors with eigenvalues > 1, as explained later)
  - Examine Scree plot, conduct a parallel analysis
  - Rerun analysis with different options and examine loadings, if the results have good simple structure
Step 3 : If more than 1 component or factor is retained, rotation of components or factors is often performed in order to obtain a more interpretable pattern of loadings. Many methods of rotation are discussed in advanced treatments of factor analysis
- Oblique rotations are best to use as they do not assume factors are NOT correlated
  - direct oblimin , quartimin, and promax
Step 4 : Items to retain
- .40 higher is often considered a minimum
  - Represents 16% of the variance in the given item (measure) is explained (associated with) the given factor
Decide which items to retain on a given factor
Eliminate items with cross-loadings
- Lots of different opinions regarding what is a cross-loading worthy of elimination (opinions and guidelines vary!)
- Rules of thumb
  - loadings of .30+ on more than one variable are worrisome
  - there should be a gap of at least .20 - .30 between the primary factor and the potentially cross-loaded factor
- Most important, clearly state your approach (cite a source for the recommendation) and stick to it!