Modeling Unobserved Heterogeneity Using Latent Profile Analysis: A Monte Carlo Simulation

Peugh, James; Fan, Xitao

doi:10.1080/10705511.2013.824780

Cited by 249 publications

(208 citation statements)

References 34 publications

Supporting

Mentioning

203

Contrasting

Order By: Relevance

“…For instance, confirming the conclusions from numerous studies (Li & Hser, 2011;Nylund et al, 2007;Peugh & Fan, 2013Tofighi & Enders, 2007), the AIC proved to be the least reliable of the ICs (with a correct class enumeration rate of 74.22 % across conditions), and the one most sensitive to design conditions. Similarly, the entropy (with a correct class enumeration rate of 69.5 % across conditions) proved to be unsatisfactory and highly sensitive to the design conditions.…”

Section: Discussionsupporting

confidence: 62%

Performance of growth mixture models in the presence of time-varying covariates

Diallo

Morin

2016

Behav Res

View full text Add to dashboard Cite

Growth mixture modeling is often used to identify unobserved heterogeneity in populations. Despite the usefulness of growth mixture modeling in practice, little is known about the performance of this data analysis technique in the presence of time-varying covariates. In the present simulation study, we examined the impacts of five design factors: the proportion of the total variance of the outcome explained by the time-varying covariates, the number of time points, the error structure, the sample size, and the mixing ratio. More precisely, we examined the impact of these factors on the accuracy of parameter and standard error estimates, as well as on the class enumeration accuracy. Our results showed that the consistent Akaike information criterion (CAIC), the sample-size-adjusted CAIC (SCAIC), the Bayesian information criterion (BIC), and the integrated completed likelihood criterion (ICL-BIC) proved to be highly reliable indicators of the true number of latent classes in the data, across design conditions, and that the sample-size-adjusted BIC (SBIC) also proved quite accurate, especially in larger samples. In contrast, the Akaike information criterion (AIC), the entropy, the normalized entropy criterion (NEC), and the classification likelihood criterion (CLC) proved to be unreliable indicators of the true number of latent classes in the data. Our results also showed that substantial biases in the parameter and standard error estimates tended to be associated with growth mixture models that included only four time points.

show abstract

Section: Discussionsupporting

confidence: 62%

Performance of growth mixture models in the presence of time-varying covariates

Diallo

Morin

2016

Behav Res

View full text Add to dashboard Cite

show abstract

“…This relatively strong assumption is frequently too stringent for real-life settings, and can result in the overextraction of spurious latent profiles (Bauer & Curran, 2003;Peugh & Fan, 2013). Notably, highlighted the unrealism of this requirement when profile indicators can be assumed to reflect a single overarching factor (i.e., global motivation in the present study, e.g., McInerney & Ali, 2006), and proposed factor mixture analyses as a way to relax this conditional independence assumption.…”

Section: The Present Studymentioning

confidence: 99%

Achievement goal profiles among adolescent males and females.

Litalien

Morin

McInerney

2017

Developmental Psychology

View full text Add to dashboard Cite

Achievement goal theory has long been a dominant model in the study of student motivation. However, a relatively small number of researchers have investigated gender differences in achievement goals or have considered the possible role that social and extrinsic goals may play in student academic motivation. Adopting a person-centered multiple goals perspective based on personal investment theory, this longitudinal study investigated whether males and females shared similar goal profiles, and whether the predictors (facilitating conditions) and outcomes (learning processes, task perseverance, and future aspirations) of these profiles were equivalent across genders. Profiles were extracted from 8 types of academic goals, based on a large sample of Hong Kong high school students (N = 7,848). Findings revealed 5 distinctive profiles for both males and females. Although the relative size of these profiles differed across samples of male and female students, the results show that 4 of these profiles were mostly equivalent across genders. Predictors of membership into these profiles were also equivalent across genders, whereas their relative outcomes were specific to gender.

show abstract

“…In mixture models, deciding how many profiles to retain is guided by an examination of the theoretical meaning and conformity of the extracted profiles Muthén, 2003) Simulation studies indicate that four of these indicators (CAIC, BIC, ABIC, and BLRT) are particularly effective and that when the indicators fail to retain the optimal model, the ABIC and BLRT tend to overestimate the number of classes, whereas the BIC, CAIC, and aLMR tend to underestimate it (Nylund et al, 2007;Peugh & Fan, 2013;Tein, Coxe, & Cham, 2013;Tofighi & Enders, 2008;Tolvanen, 2007;Yang, 2006). However, these tests are variations of tests of statistical significance, heavily influenced by sample size , so that with sufficiently large sample sizes, these indicators may keep on suggesting the addition of profiles without ever reaching a minimum.…”

Section: Analysesmentioning

confidence: 99%

“…All reported models converged on a replicated solution and can be assumed to reflect a "real" maximum likelihood. For each parameterization, models with 1 to 8 latent profiles were estimated with the indicators' (wellbeing factor scores) intercepts and residuals freely estimated in all profiles Peugh & Fan, 2013). However, for FMA (Model 3), these models converged on improper solutions (negative variance estimates, non-positive definite Fisher Information matrix, etc.)…”

Section: Analysesmentioning

confidence: 99%