Get startedGet started for free

Model the WAIS-III IQ Scale

1. Model the WAIS-III IQ Scale

Now that you have practice working with lavaan, we will use these skills to examine a recent model of intelligence, namely the Wechsler Adult Intelligence Scale version III.

2. WAIS-III Picture

The WAIS-III is a four-factor model of intelligence. First, verbal comprehension measures how well a person can listen and respond to tasks. For vocabulary, a person would define words, while the similarity task requires a person to relate concepts together. Information and comprehension cover general knowledge and cultural rules. Working memory measures how well a person can attend to and process information. These three tasks: arithmetic, digit span, and letter number sequences, require a person to remember information in a specific order. Perceptual organization measures the person's ability to examine information without words by solving puzzle designs and pictures. Picture completion, block design, and matrix reasoning are all tasks that require a person to put together a visual picture or story. Processing speed measures how quickly a person can process visual information. Lastly, digit symbol coding and symbol search are speeded visual tasks that require a person to find and copy objects.

3. WAIS-III Model

This proposed model of the WAIS is considered hierarchical and has two sets of latent variables. First, we have the four latent variables that are measured by the 12 subscale manifest variables. These four latent variables are then predicted by a second set of latent variables, which creates a two-step model or two-set hierarchy. Hierarchical models are often used when explaining very strong correlations between sets of latent variables, rather than just using covariances between all of the variables.

4. How to Get Started

As you improve your skills, you can create more complex models in lavaan. However, it is always beneficial to always start with the base level examining the relationship of the manifest variables to the latent variables. We should look for problems with the model, mainly Heywood cases, and fix these issues before adding more complexity.

5. Let's practice!

Let's try building a four-factor model of the WAIS-III subscales and see how that model fits the data.