Bayesian Shrinkage for Functional Network Models, with Applications to Longitudinal Item Response Data

Longitudinal item response data are common in social science, educational and psychology, among other disciplines. Studying the time-varying relationships between items is crucial for assessment or designing marketing strategies from survey questions. Although dynamic network models have been widely developed, we cannot apply them directly to because there multiple systems of nodes with various types local interactions items, resulting multiplex structures. We propose a new model study these temporal by embedding functional parameters within exponential random graph framework. Inference on such difficult likelihood functions contain intractable normalizing constants. Furthermore, number grows exponentially as increases. Variable selection not trivial standard shrinkage approaches do consider trends parameters. To overcome challenges, develop novel Bayes approach combining an auxiliary variable MCMC algorithm recently-developed method. our review sets, illustrating that proposed can avoid evaluation constants well detection significant items. Through simulation under different scenarios, examine performance algorithm. Our method is, knowledge, first attempt select variables