Short Courses For IMPS 2020

 

Model implied instrumental variable (MIIV) methods using MIIVsem: An R package for Structural Equation Models

Instructors: Ken Bollen & Zack Fisher

If our models were perfectly valid and our variables only came from normal distributions, the maximum likelihood and related estimators that dominate SEM software would be hard to beat. In reality, however, such structural and distributional assumptions are rarely if ever satisfied. This workshop will discuss more robust estimators that better represent real world conditions. Model Implied Instrumental Variable (MIIV) estimators are more robust to the approximate nature of our models and are asymptotically distribution free. In addition, they can test equation level fit so as to better localize model misspecification. The workshop will give an overview of the free R package MIIVsem. We will introduce the key ideas behind MIIV estimation; we will show how MIIVsem automates the selection of MIIVs, the estimation of coefficients and standard errors, and provides overidentification tests for equations. These and other features will be introduced and illustrated with a variety of empirical examples. We will provide instructions on downloading R and MIIVsem and will ask participants to bring their computers to the workshop so that they can run these empirical examples on their own machines during the last part of the workshop. No prior knowledge of R or MIIVsem is assumed.

 

Statistical learning methods for process data

Instructors: Qiwei He, Jingchen Liu, & Xueying Tang

This short course introduces several recent advancements in the analysis of process data collected from computer-based interactive items. Methods that extract information from both observed sequential actions (e.g., n-grams) and latent variables will be presented. Covered topics include (1) automatic feature extraction from action sequences and their timestamps via n-grams, multidimensional scaling, and sequence-to-sequence autoencoders, (2) supervised prediction of item response outcomes and external variables based on process data via recurrent neural nets, and (3) action sequence prediction and segmentation based on neural language models. The ProcData R package, an open source package for exploratory process data analysis, will be introduced. During the full-day short course, participants will be provided with an overview of process data collected from computer-based large-scale assessments, learn about various approaches to analyzing and using log data, and obtain hands-on experience working with log data through examples and exercises. Intended audience are researchers and practitioners with basic knowledge of latent variable modelling and familiarity with R/RStudio, who are interested in learning or applying data-driven methods to log data analysis. Running the ProcData package requires installation of R, Rcpp, and Anaconda (installation instructions/help will be provided). Participants are expected to bring their own laptop with Windows or Mac operating system.

 

Introduction to flexMIRT® version 3.6

Instructors: Li Cai, Michael Edwards, & RJ Wirth

flexMIRT® is an IRT software package for multilevel, multidimensional, and multiple group item response modeling. flexMIRT® Version 3.6 offers substantially upgraded statistical modeling capabilities, including new model fit indices, multidimensional models with non-normal latent densities, Markov chain Monte Carlo (MCMC) estimation, crossed random effects, and improved diagnostic classification modeling. This full-day training session will introduce users to the flexMIRT® system and provide hands-on experience with the software. The workshop is intended for graduate students, researchers, and practitioners in educational, psychological, and clinical outcomes assessment fields interested in the application of modern psychometrics to help improve the precision and quality of measurement. Attendees will receive a free two-month trial version of flexMIRT. It is assumed that attendees are already familiar with the basic tenets of item response theory. It would be helpful if the attendees could bring their own PCs that could run Windows 7 or above.

 

Practical causal analysis and effect estimation with observational data

Instructor: Yiu-Fai Yung

Due to practicality and limited resources for doing randomized experiments, psychological research often needs to make causal inferences from observational data. But when does an effect estimate have a valid causal interpretation? This course introduces commonly used methods for estimating dichotomous treatment effects from observational data and graphical tools for evaluating the conditions under which the effect estimate has a valid causal interpretation. Specifically, this course discusses estimation methods such as propensity score matching, inverse probability weighting, and doubly robust methods. It reviews the regression methods for causal mediation analysis. In addition, it describes the role of directed graphs as a tool to represent the data generating process, communicate and incorporate expert knowledge, reason about sources of association and bias, and construct a valid estimation strategy. After a concise review of the theory behind these estimation and graphical methods, simulated real-world examples illustrate their applications by SAS® software. These examples demonstrate good practices and effective strategies for dealing with practical challenges. This course emphasizes analytical techniques for applications but not the software per se. Knowledge learnt from this course should transfer to applications with other software. No prior experience with causal modeling is required.