Extending the Functionalities of {metafor}

[Info]
This event is a follow-up from the 26 August 2021 Oslo useR! event. You can find the recording of August event here: https://youtu.be/IkduL5iRdqo

[Talks]

Meta-Analysis of Dependent Effect Sizes: Robust Variance Estimation with {clubSandwich} (James Pustejovsky)

Large meta-analyses often involve dependent effect sizes, but where the exact form of the dependence is unknown. Meta-analysis with robust variance estimation handles this problem through specification of a working model for the dependence, which need not be correct. However, the two currently available working models are limited to each describing a single type of dependence. James will demonstrate a workflow for implementing an expanded set of working models by combining the metafor and clubSandwich R packages. A pre-print is available at https://osf.io/preprints/metaarxiv/vyfcj/.

Introduction to Meta-Analytic Structural Equational Modeling with {metaSEM} (Diego G. Campos)

We often formulate models to understand how our data is connected. However, it is difficult to assess whether our model is a good representation of the data or the generalizability of our model to other contexts. With {metaSEM}, researchers can combine data from several primary studies to investigate the generalizability of a model. In this talk, we will show you the steps to perform a meta-analytic structural equational modelling (MASEM) with the {metaSEM} package.

Calculating the Statistical Power of Studies Included in a Meta-Analysis Using {metameta} (Daniel Quintana)

Several methods exist for assessing study quality in meta-analysis, but these are typically discipline specific. One approach used across disciplines to assess the evidential value of an individual study is to calculate its statistical power, which can identify the range of effect sizes that can be reliably detected. Studies that cannot detect realistic effect sizes are unlikely to be replicated. A meta-analysis with mostly studies not capable of reliably detecting realistic effects would therefore have less credibility. This talk describes {metameta}, an R package that can effortlessly calculate and visualise the statistical power of studies included in a meta-analysis.

Meta-Analysis of Nonparametric Models with {metagam} (Øystein Sørensen)

"Most meta-analytic tools have focused on parametric statistical models, and software for meta-analyzing nonparametric and semiparametric models like generalized additive models (GAMs) have not been developed. The metagam package attempts to fill this gap: It provides functionality for removing individual participant data from GAM objects such that they can be analyzed in a common location, as well as various tools for visualization and statistical analysis."

[Speakers]

James Pustejovsky is a statistician and Associate Professor in the School of Education at the University of Wisconsin-Madison. His research involves developing statistical methods for problems in social science research, with a focus on meta-analysis methods. He completed his PhD in statistics from Northwestern University in 2013. His homepage is https://www.jepusto.com/

Diego G. Campos is a doctoral research fellow at the Centre for Educational Measurement at the University of Oslo. His research covers meta-analysis, structural equational models, and multilevel analysis. Diego is currently working on the use of IPD meta-analysis for the synthesis of complex survey data.

Daniel Quintana (@dsquintana) is a Senior Researcher in the Department of Psychology at the University of Oslo and co-host of the research methodology podcast "Everything Hertz".

Øystein Sørensen is an Associate Professor in Statistics at the Department of Psychology, University of Oslo. He works on both applied data analysis projects as well as the development of new statistical tools and is the author of four R packages. He received his Ph.D. from the UiO in 2014. His homepage is https://osorensen.rbind.io/