Explore chapters and articles related to this topic
Latent Profile Analysis
Published in Douglas D. Gunzler, Adam T. Perzynski, Adam C. Carle, Structural Equation Modeling for Health and Medicine, 2021
Douglas D. Gunzler, Adam T. Perzynski, Adam C. Carle
Latent profile analysis (LPA) relates a set of observed continuous indicators to a set of latent profiles (underlying subgroups). Broadly, LPA differs from LCA (Chapter 11) in using continuous indicators as opposed to categorical indicators. For example, suppose a researcher was interested in assessing whether scores on a set of clinical outcome measures, such as systolic blood pressure, diastolic blood pressure, body mass index, triglycerides, high-density lipoprotein and hemoglobin A1c tended to cluster together in specific ways. There are patients that do poorly on all of these outcomes or well on all of these outcomes, whereas others do well on some but not others. Then, LPA would be appropriate.
Latent Transition and Growth Mixture Models
Published in Jason T. Newsom, Longitudinal Structural Equation Modeling, 2015
A second model was tested to illustrate a latent profile analysis with continuous indicators. For this model, the health and aging data set was used so that the results could be presented for the latent transition model described below, a model that requires more time points than available from the social exchange data set. The latent profile model estimated positive affect classes using five continuous indicators, frequency ratings for the “happy,” “enjoy,” “satisfied,” “joyful,” and “pleased” questions. An initial model specified three classes. No chi-square is available for models with continuous indicators. The aBIC was equal to 5,252.331, but this value can only be used to judge relative fit. As a comparison, the aBIC for a two-class model with the same indicators was 6,575.164, a considerably higher value that indicates the two-class model had a poorer fit. The likelihood ratio tests for comparing the fit of this model to a model with two classes, suggested that the three-class model fit the data significantly better, LMR=170.942, p < .001, VLMR=176.372, p < .001. The mean of the first class was .202 and the mean of the second class was 1.315. Using the cumulative normal cdf transformation, the corresponding class membership proportions obtained were .206, .626, and .168 for the three classes. Table 10.1 presents the within-class intercepts for the five indicators. The observed means presented in the table are for a 5-point scale of agreement, with 5 representing the most positive. The intercepts appear to increase across classes, with the lowest positive affect rating in the first class and the higher positive affect ratings in the third class. This pattern seems to indicate that there are three groups of respondents, unhappy individuals, moderately happy individuals, and very happy individuals, though further investigations would be needed to obtain more information about the meaning of these classes.
Understanding sexual and gender minority substance use through latent profiles of ecological systems
Published in Substance Abuse, 2022
Barrett Scroggs, Heather A. Love
One of the key findings from the latent profile analysis was the distribution of participants in the identified profiles. Half (48%, n = 886) of the participants in our sample fit into the moderate with low self-esteem profile. It is interesting to note that this profile was associated with more frequent amphetamine and heroin use compared to the moderate with high self-esteem profile. As both of these profiles had moderate levels of all other ecological experiences (hope, community connection, and ACEs), the ecological experience that arguably put them at risk was their low self-esteem, despite otherwise positive experiences. This key difference supports the justification for the 4-profile solution as opposed to the 3-profile solution that was explored. Additionally, these differences in self-esteem highlight a potential intervention area for preventing increased use of amphetamine and heroin use in SGM individuals. Practitioners should acknowledge the likelihood that SGM emerging adults may be dealing with lower levels of self-esteem, potentially linked to the minority stress they experience living in a heteronormative and cis-normative society that often diminishes or outright rejects them. As identified in previous literature,25 self-esteem appears to be a significant protective factor against engaging in riskier substance use
Distinct Latent Profiles of Working Memory and Processing Speed in Adults with ADHD
Published in Developmental Neuropsychology, 2021
Sophie I. Leib, Richard D. Keezer, Brian M. Cerny, Lindsey R. Holbrook, Virginia T. Gallagher, Kyle J. Jennette, Gabriel P. Ovsiew, Jason R. Soble
Latent Profile Analysis (LPA) was conducted in Mplus Version 8.5. Latent profile analysis is a person-centered statistical approach used to classify individuals from a heterogeneous population into smaller more homogenous subgroups based on probability of group membership (Berlin, Williams, & Parra, 2014; Muthen, 2001). No explicit sample size recommendations for LPA are available, but recent work suggests that a sample size of >100 is sufficient for detecting medium effects (Dziak, Lanza, & Tan, 2014). Therefore, our current sample of 179 was deemed appropriate for analysis. ACSSs for the two working memory tasks (Digit Span and Letter Number Sequencing) and the two processing speed tasks (Symbol Search and Coding) were used as profile indicators to identify latent groups of performance.
Who are the Adolescents Admitted to an Inpatient Unit? Results of a Latent Profile Analysis from an Acute, Psychiatric Hospital
Published in Evidence-Based Practice in Child and Adolescent Mental Health, 2021
Natalie Rodriguez-Quintana, Ana M. Ugueto
After identifying the profile membership, one set of six linear regression models were run to determine if there were statistically significant differences between the profiles and the variables used in the LPA (e.g., is there a statistically significant difference in CES-DC scores between profiles). Another set of five regression models were run to determine if there were statistically significant differences between the six obtained profiles and demographic (i.e., age, race/ethnicity, gender, and length of stay) and suicide risk variables. Adolescent profile was considered the independent variable in the linear regression models. TidyLPA for R was used for latent profile analysis (Rosenberg et al., 2018) and IBM SPSS 26 (IBM Corp., 2019) was used for linear regression analyses.