Methods Festival 2017Session description form
30.-31.5. University of Jyväskylä(Speakers fill in and return to their
contact person)
Session name: Multilevel mixture models
Session Chair: Asko Tolvanen
Session content and objective:
Themes and speakers:
Name: Asko Tolvanen
Title of the presentation: General about multilevel mixture modelling
Mixture modelling in the context of multilevel/hierarchical data offers new very interesting possibilities to analyze data. As in single level data there are many different basic models (regression, path,and factor models) in which parameters of the latent classes can differ. In this presentation different steps and aspects to build mixture model is discussed. This presentation works as introduction to two successive presentations of special cases of multilevel mixture modeling.
Name: MinnaTorppa
Title of the presentation: Multilevel factor mixture analysis
Description of the presentation: This paper presents a multilevel factor mixture analysis (FMA) example. The data comes from the Jyväskylä Longitudinal Study of Dyslexia (n= 1750 from 93 classrooms). Data is longitudinal with four measurement points during the first two school years. The purpose of the study was to examine whether heterogeneous developmental paths can be identified based on profiles of word recognition and reading comprehension (multilevel FMA). Because the data comes from classrooms, we had to examine and control for the classroom membership effect by using multi-level modelling. Secondly, we studied what kind of early language and literacy skill profiles and reading experiences characterize the children with differing reading development in a smaller sub-sample. The mixture modeling procedure resulted in five subtypes: (1) poor readers, (2) slow decoders, (3) poor comprehenders, (4) average readers, and (5) good readers. Group differences were found in the early language and literacy skill development as well as in the reading experiences of the reading subtypes. In the presentation I will describe the analysis procedure and explain the reasons underlying methodological choices and decisions.
Name: Anne Mäkikangas
Title of the presentation: Multilevel latent profile analysis
Description of the presentation: Latent Profile Analysis (LPA) is a person-centered method commonly used in organizational research to identify homogeneous subpopulations of employees within a heterogeneous population. However, in the case of nested data structures, e.g., employees nested in departments, multilevel techniques are needed. Multilevel LPA (MLPA) enables adequate modeling of subpopulations in hierarchical datasets. Besides accurate Level 1 profile classifications, MLPA enables investigation of a) the similarity in the proportional distribution of Level 1 profiles across Level 2 units; b) Level 2 latent profiles based on the proportions of the Level 1 latent profiles; and c) the extent to which covariates drawn from the different hierarchical levels affect the probability of a particular profile. We demonstrate these advantages of MLPA by investigating job strain patterns using data from 1,958 university employees clustered in 78 work departments. The implications of the results for organizational research are discussed, together with several issues related to applying MLPA and its possibilities for wider application.