During the pre-conference a one day program is presented to JURE members of the EARLI SIGs 18 and 23 attending the main conference. Registration for the pre-conference can be done simultaneous with registration for the main conference. Registration opens on April 24th 2018.
A tentative program timeline for the pre-conference is presented below. During this day there are two small group workshops of three hours from which the participants can choose. The choice of workshop has to be entered in the registration for the conference. For each workshop a maximum of 20 participants is set.
The first workshop, provided by Jessaca Spybrook, is focused on designing adequately powered cluster randomized trials. The second workshop, provided by George leckie, is focused on three-level, cross-classified and multiple membership multilevel models. See below for the full workshop descriptions.
The pre-conference will be held at the Faculty of Behavioural and Social Sciences of the University of Groningen. Please see check the pre-conference venue page (http://www.earli2018sig1823.nl/pre-conference-venue/) for more information on this venue and how to get there.
JURE Pre-conference (tentative)
|10.00 – 10.30||Welcome and coffee|
|10.30 – 11.00||Opening|
|11.00 – 12.30||Workshop 1 or 2 (part 1)|
|12.30 – 13.30||Lunch|
|13.30 – 15.00||Workshop 1 or 2 (part 2)|
|15.00 – 15.30||Coffee break|
|15.30 – 17.00||Meet the professors|
|Optional having dinner together|
Designing Adequately Powered Cluster Randomized Trials: A Hands-On Workshop
Jessaca Spybrook, PhD
Western Michigan University
Cluster randomized trials, for example studies with students nested within schools with treatment randomly assigned at the school level, are becoming common designs to test the effectiveness of educational interventions. In order to yield high-quality evidence of the effectiveness of the intervention, the study must be designed with adequate statistical power to detect a meaningful treatment effect. The purpose of this workshop is to describe statistical power calculations for the main effect of treatment for 2-level CRTs. Further, I will introduce power calculations for moderator effects, or differential treatment effects for 2-level CRTs. The focus will include the conceptual framework for the calculations as well hands-on practice with the free software programs PowerUp! and PowerUp!-Moderator.
Multilevel Modelling: Three-level, Cross-Classified and Multiple Membership Models
Dr George Leckie
Reader in Social Statistics
Centre for Multilevel Modelling, School of Education, University of Bristol
Multilevel modelling is now standard in quantitative educational research, but most researchers continue to only consider two-level analyses even when this is inappropriate. This intermediate course moves beyond standard two-level analyses to consider multilevel models for data with more complex three-level, cross-classified, and multiple-membership structures. Like two-level models, these more advanced models can be viewed as extended linear regressions where the intercept and regression coefficients are all potentially allowed to vary across clusters. However, we now have multiple sources of clustering and this introduces new model specification, estimation, software, and interpretation challenges. In this course, we will introduce each class of model and their challenges and we will illustrate them with applications which extend standard two-level students-within-schools analyses of continuous student achievement. Thus, in our three-level application we will additionally account for clustering by classrooms, recognizing that students from the same classroom have more similar outcomes than students from different classrooms. In our cross-classified application we will instead account for students’ residential neighbourhoods, recognising that not all students from the same neighbourhood attend the same school; a three-level model is not appropriate. In our multiple membership application we will allow for pupil mobility between schools and will therefore respect the entire sequence of institutions attended, not just the final one. We will provide electronic materials so that participants can replicate the presented analyses after the course in the MLwiN and Stata statistics packages.