Research

Starting in early 2022, FAIR will contribute to three current research areas in the social sciences: prediction modeling, agile intervention design and evaluation, and targeted methodological research.

Prediction Modelling

In this area, we use innovative methods from the data sciences to analyze large-scale data, in particular, from IGLU and NEPS. Our research with IGLU data focuses on the prediction of reading competence. Our research with NEPS data focuses on educational and career choices related to math-intensive fields. Additional data sets are considered as well (e.g., BONNS; Test-M-I).  One cornerstone of FAIR's research on aging and well-being is SHARE's data, especially the SHARELIFE subset.

We plan to reevaluate existing data e.g. with predictive regression models using sketching and coreset methods to improve their scalability. The flexibility of sketching and the Merge & Reduce framework will enable applications on streamed or dynamic data updates (e.g. for analyzing sensor data or behavioral traces). 

Targeted Methodological Research

FAIR PIs have previously developed tailor-made methods in fields like biostatistics. These current methodological innovations have yet to find their way into the social sciences, because the novel methods have good inference properties for heterogeneous experimental settings or in situations with small sample sizes (e.g., rare subgroups). Methods can also handle complex situations such as missing values, high dimensions or unbalancedness. FAIR will systematically adapt and extend these methods for social science contexts covering; e.g., repeated measures and complex panel data. Moreover, tree-based prediction models and random effects regression will be used for the development of new prediction models.

To facilitate the analysis of large and high dimensional FAIR will adapt and further develop so called sketching and coreset methods. Those can be used to reduce data to an approximate, but still sufficient statistic of significantly smaller size while retaining the statistical power of the full data. Those methods will be adapted and transferred for their use in social science applications (e.g. for large scale panel data).

Agile Intervention Design and Evaluation

Key goal in the intervention study is to investigate the effects of two different, tablet-based interventions that that aim to improve mathematical skills in first-grade elementary school children. The first intervention focuses on visual perception and patterning, the second intervention targets cardinal and relational number knowledge. By conducting this study, we aim to dissociate the effects of visual perception and visual skills from the effects of numerical cognition on the development of mathematical competencies.