Ms. Bai is a dual - title Ph.D. candidate in Educational Theory & Policy and Comparative International Education at the Pennsylvania State University specializing in a variety of statistical methods,
including multilevel modeling, structural equation modeling, and propensity score matching.
Not exact matches
Examples of his contributions
include improved effect size estimates,
multilevel mediation
models, and Bayesian approaches to mediation analysis.
His methodological research interests
include item writing, reliability theory and
multilevel item response
models, with a substantive interest in Latino youth development.
Perhaps the most enabling resource for the educational research community was Singer's (1998) article demonstrating how to implement
multilevel (
including growth)
models using one of the most widely available general - purpose statistical packages.
Rather, I recommend that they --- you --- become aware (to the best of your technical ability) of how these methods work, so you can use them in cases where they are most appropriate (these situations would
include forecasting,
multilevel modeling, inference for complex
models with many parameters, and settings with weak data).
Articles discuss methodological challenges and opportunities in family and couple research,
including outcome, cost - effectiveness, qualitative, and narrative research; video - recall procedures,
multilevel methods, diary methods, and cluster analysis; and moderator effects, the actor — partner interdependence
model, survival analysis, and ethical issues.
Multilevel models are ideal because families with missing data are not removed from the analyses, and the
models are not affected by unbalanced data (i.e., in which some families
include more member reports than others).
Next, because the data
included multiple reports from individuals within the same family, we used
multilevel modeling to address the issue of dependencies in the data (SAS PROC MIXED; Singer, 1998).
Multilevel modeling was also conducted on each outcome, with condition, time, and the condition × time interaction
included in the
model; random intercepts and slopes were estimated for each participant.
Two
multilevel models were estimated, one without baseline functioning and one
including baseline outcome variables when they were collected with the first
multilevel model similar to the analysis conducted in Sure Start.
Second, because baseline data were available for many outcomes, a second
multilevel model was run that
included baseline functioning.
Control variables in the
multilevel modeling and multiple - linear regression analyses
included gender, race, and pretest scores on the outcome being predicted.
Thus, a posterior distribution was obtained for this difference, and the 95 % credible interval of this difference was then used as a decision criterion: When 0 was
included in the credible interval, there was no evidence that there are two different mean inertias, so we selected the
multilevel AR
model; when 0 was not
included in the credible interval of the mean difference, this was taken as evidence that there are two distinct states with different mean inertias, so we selected the
multilevel TAR
model.
First we conducted an additional analysis in the
multilevel models that
included a four - category couple drinking variable and gender as well as the interaction between gender and couple drinking categories as the predictors.
First, with
multilevel modeling, one is able to take advantage of all available data,
including data from participants who did not complete all assessments.
This general approach — to first quantify the intradyad relationships and then examine interdyad differences in the intradyad relationships — is the basis for most contemporary dyadic data analysis techniques,
including sequential and state space grid analyses, coupled dynamic systems, and
multilevel modeling (Bakeman & Gottman, 1997; Bakeman & Quera, 2011; Boker & Laurenceau, 2007; Gonzalez & Griffin, 2012; Gottman, Murray, Swanson, Tyson, & Swanson, 2002; Hollenstein, 2013; Laurenceau & Bolger, 2005; Ram & Pedersen, 2008).