Local solutions in the estimation of growth mixture models.

By John R. Hipp, Daniel J. Bauer. Psychological methods, 2006

717 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Statistics and Probability

View Paper: https://www.semanticscholar.org/paper/011542426b43b4383b4112150f09d95329e19e87

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Statistics and Probability

Citations in the corpus

Mixture Modeling for Lifespan Developmental Research (2019, background): “Technically, mixture models are estimated through an iterative process that risks converging on a local solution rather than on a true maximum likelihood when starts values are not adequate (Hipp & Bauer, 2006).”
Stability, change, and implications of students’ motivation profiles: A latent transition analysis (2017, methodology): “To avoid local maximum, all LPA were conducted using 5000 random sets of start values, 1000 iterations, and retained the 200 best solutions for final stage optimization (Hipp & Bauer, 2006; McLachlan & Peel, 2000).”
Disentangling Shape from Level Effects in Person-Centered Analyses: An Illustration Based on University Teachers’ Multidimensional Profiles of Effectiveness (2015, methodology): “All analyses were conducted with Mplus 6.1 (Muthén & Muthén, 1998–2010), using the robust maximum likelihood estimator (MLR) and 2,000 random starts, 100 iterations for these random starts, and the 100 bests retained for final stage optimization (Hipp & Bauer, 2006; McLachlan & Peel, 2000).”
General Growth Mixture Analysis with Antecedents and Consequences of Change (2010, positive): “The use of multiple starts from random locations in the parameter space can improve chance of convergence to global maxima (Hipp & Bauer, 2006).”
Emotional labour profiles: Associations with key predictors and outcomes (2018, methodology): “All LPA were conducted using 5000 random sets of start values, 1000 iterations, and retained the 200 best solutions for final stage optimisation (Hipp & Bauer, 2006).”
LATENT GROWTH MIXTURE MODELING: A SIMULATION STUDY (2007, background): “It is a known feature that the LGMM estimation may often stop to the local maximum of log-likelihood producing biased parameter estimates (Hipp & Bauer, 2006).”
Methods and Measures: Growth mixture modeling: A method for identifying differences in longitudinal change among unobserved groups (2009, background): “Given that the solutions obtained during model estimation may sometimes not be accurate or useful (cf. Hipp & Bauer, 2006), it is important to first check that the model makes sense mathematically.”
Are Commitment Profiles Stable and Predictable? A Latent Transition Analysis (2016, background): “We also increased the default to 100 iterations for these random starts and retained the 100 best solutions for final stage optimization (Hipp & Bauer, 2006; McLachlan & Peel, 2000).”
A Multifoci Person-Centered Perspective on Workplace Affective Commitment: A Latent Profile/Factor Mixture Analysis (2011, methodology): “It is recommended to estimate the model with multiple random sets of start values (Hipp & Bauer, 2006; McLachlan & Peel, 2000).”
CHAPTER 25 Latent Class Analysis and Finite Mixture Modeling (2013, neutral): “…values with the estimation algorithm (it is recommended that a minimum of 50-100 sets of extensively, randomly varied starting values be used (Hipp & Bauer, 2006}, but more may be necessary to observe satisfactory replication of the best maximum log likelihood value} and keep track of the…”
Observations on the Use of Growth Mixture Models in Psychological Research (2007, background): “individual (and mean) trajectories is assumed to be equal such that ƒk D ƒ. The dispersion matrices may also be held equal over classes, that is, ‰k D ‰ and ‚k D ‚. Seldom motivated by theory, these constraints are most often imposed for statistical expedience (e.g., to guarantee a global maximizer;”
An introduction to latent class growth analysis and growth mixture modeling. (2008, background): “A third issue that is often raised is the problem of nonconvergence and local solutions (cf. Hipp & Bauer, 2006).”
Is creativity domain-specific? Latent class models of creative accomplishments and creative self-descriptions. (2009, neutral): “Unlike SEM, latent class models have many local solutions on the likelihood surface (Hipp & Bauer, 2006).”
Multiple-Group Analysis of Similarity in Latent Profile Solutions (2016, methodology): “The simpler country-specific models were estimated using 5,000 random sets of start values with 100 iterations and the 200 best solutions retained for final stage optimization (Hipp & Bauer, 2006).”
Modeling Unobserved Heterogeneity Using Latent Profile Analysis: A Monte Carlo Simulation (2013, background): “A well-known difficulty with LCAs and LPAs is the tendency of the maximum likelihood (ML) parameter estimation algorithm to converge on a local maxima rather than the true maximum of the ML fit function, or fail to converge (e.g., Hipp & Bauer, 2006; McLachlan & Peel, 2000).”
Latent profile analysis: A review and “how to” guide of its application within vocational behavior research (2020, methodology): “To guard against local solutions, we increased the default settings in M plus and set the number of random starts to 7000 and the final stage optimizations to 200 (Hipp & Bauer, 2006).”
Finding latent groups in observed data: A primer on latent profile analysis in Mplus for applied researchers (2020, methodology): “To test whether the model resulted from a local maxima in the estimation process, different, random start values can be used to run the model again (see Hipp & Bauer, 2006 for further information).”
Complementary Variable- and Person-Centered Approaches to the Dimensionality of Psychometric Constructs: Application to Psychological Wellbeing at Work (2016, methodology): “2 (Muthén and Muthén 2014) MLR estimator, using 10,000 random starts, 1000 iterations for these random starts, and the 500 best retained for final stage optimization (Hipp and Bauer 2006; McLachlan and Peel 2000).”
Classical Latent Profile Analysis of Academic Self-Concept Dimensions: Synergy of Person- and Variable-Centered Approaches to Theoretical Models of Self-Concept (2009, neutral): “We are thus reasonably confident that these solutions represent the best fitting solution, but it should be noted that this potential problem of local maxima represents a potentially serious problem for LPAs and mixture models more generally (see Hipp & Bauer, 2006).”