Factorization of person response profiles to identify summative profiles carrying central response patterns.

Abstract

A data matrix, where rows represent persons and columns represent measured subtests, can be viewed as a stack of person profiles, as rows are actually person profiles of observed responses on column subtests. Profile analysis seeks to identify a small number of latent profiles from a large number of person response profiles to identify central response patterns, which are useful for assessing the strengths and weaknesses of individuals across multiple dimensions in domains of interest. Moreover, the latent profiles are mathematically proven to be summative profiles that linearly combine all person response profiles. Since person response profiles are confounded with profile level and response pattern, the level effect must be controlled when they are factorized to identify a latent (or summative) profile that carries the response pattern effect. However, when the level effect is dominant but uncontrolled, only a summative profile carrying the level effect would be considered statistically meaningful according to a traditional metric (e.g., eigenvalue ≥1) or parallel analysis results. Nevertheless, the response pattern effect among individuals can provide assessment-relevant insights that are overlooked by conventional analysis; to achieve this, the level effect must be controlled. Consequently, the purpose of this study is to demonstrate how to correctly identify summative profiles containing central response patterns regardless of the centering techniques used on data sets.

Examination of COVID stress syndrome facets and relations to substance misuse using profile analysis via multidimensional scaling (PAMS)

“…First, it involves interpreting dimensions as patterns (latent profiles) rather than factors (latent variables), which represent the major patterns of individual profiles included in a sample. Second, it produces profiles via modeling at the individual level, which permits interpreting a person's scores with regard to the extent to which they align with one or more of the profiles extracted (Kim, 2023;Kim et al, 2007). Latent profile analysis (LPA) has a similar function as PAMS, but a latent profile of LPA is conceptualized as a list of the average subscale scores for each latent class identified by LPA (Liu et al, 2022).…”

“…Latent profile analysis (LPA) has a similar function as PAMS, but a latent profile of LPA is conceptualized as a list of the average subscale scores for each latent class identified by LPA (Liu et al, 2022). On the other hand, PAMS identifies a set of latent profiles from a sample consisting of person profiles, The PAMS latent profiles contain a subscale mean profile as well as unique latent profiles that show central response pattern information, which is not available in LPA latent profiles (Kim, 2023). In the present study, using a sample entirely unique from previous CSS research, our aims were twofold: 1) elucidate the most typical profiles of the COVID stress syndrome using PAMS, and 2) examine the relationship between the profile(s) and measures of behavioral and cognitive aspects of alcohol and drug misuse.…”