A note on using eigenvalues in dimensionality assessment

The dimensionality of a set of items is important for scale development. In practice, tools that make use of eigenvalues are often used to assess dimensionality. Parallel analysis is featured here as it is becoming an increasingly popular method for assessing the number of dimensions, and computational tools have recently been made available which will likely increase its use by practitioners. The current paper argues that methods that use eigenvalues to ascertain the number of factors may perform poorly under certain conditions, particularly for increasing levels of variable complexity and/or inter-factor correlations in the latent structure. A simulation study and an example are offered to substantiate this assertion.