Group Iterative Multiple
Model Estimation (GIMME)

Person-specific modeling allow us to better understand dynamic processes for individuals. However, false positives and missed relations can often occur when doing this type of analyses. This workshop focuses on one method, GIMME, that has been shown to reliably recover true relations for individuals as well as detect which relations consistently replicate across individuals so that generalizable inferences can be made.

Instructor:
Kathleen Gates (University of North Carolina - Chapel Hill)

Workshop Dates and Times:
Thursday, June 13th, 2024, 10:00am to 3:30pm ET
Friday, June 14th, 2024, 10:00am to 3:30pm ET

Workshop Format:
Two-Day Synchronous Online Workshop

The relationships among individuals' behaviors, emotions, and physiology are complex, and accurately quantifying these processes requires the right tools. This workshop will teach you how to analyze time series data (at least 60 time points) across multiple participants using GIMME, one method that reliably recovers the true relations among human constructs and measures. GIMME provides individual-level estimates which allows for person-specific inferences and testing of individual differences. Additionally, it identifies which relations replicate across individuals for robust generalizable or population-level inferences.

This workshop teaches the basics of the GIMME algorithm as well as extensions. Specific topics are: time series models; search procedure done in GIMME to find relations; post hoc probing of output; arriving at data-driven subsets of individuals using unsupervised classification; using latent variable modeling with GIMME; and how to best model exogenous variables (e.g., task or intervention onset; weather). Published results across various domains of inquiry are presented as examples.

Hands-on data examples using simulated data will be used throughout the workshop alongside lecture. Participants are welcome to use their own data. While all necessary code is provided, some familiarity with R will be helpful.

What you’ll learn

  • Multivariate time series models – The GIMME algorithm is flexible in terms of the model used. We'll introduce the basics of vector autoregression and structural vector autoregression.

  • GIMME algorithm – The GIMME algorithm can be applied to numerous estimation frameworks. You will learn this algorithm and how it has been applied in the freely distributed packages, as well as how you might apply it outside of these.

  • Available extensions – Many popular extension exist, with the most commonly used one being unsupervised classification of individuals based on their dynamic processes. All GIMME extensions will be discussed with code provided.

  • Coding – We'll conduct analyses using examples. Participants are welcome to use their own data.

Syllabus

Day 1: GIMME Fundamentals

  1. Why use GIMME?

  2. Vector autoregression models

  3. GIMME algorithm

  4. Data preparation

  5. Conducting GIMME

Day 2: GIMME Extensions

  1. Probing output

  2. Data-driven subgroups

  3. Confirmatory subgroups

  4. Hybrid SVAR model

  5. Exogenous variables

  6. Latent variables

  7. Setting confirmatory paths

Registration Options

Group Iterative Multiple Model Estimation

  • Professional
  • $599
  • Baseline Price for Faculty,
    Staff, and Other Professionals
  • Click Register Below
  • Trainee
  • $599 $399
  • 33% Discount for
    Students and Postdocs
  • Use code "TRAINEE" at Checkout

Combo 1: Group Iterative Multiple Model Estimation + Time Series Analysis for Intensive Longitudinal Data

  • Professional
  • $1198 $959
  • Baseline Price for Faculty,
    Staff, and Other Professionals
  • 20% Combination Discount
  • Click Register Below
  • Trainee
  • $959 $639
  • 33% Discount for
    Students and Postdocs
  • 20% Combination Discount
  • Use code "TRAINEE" at Checkout

 FAQs

  • This workshop is intended for people who have (or plan to have) time series data and wish to make person-specific or generalizable inferences. The data can be from ecological momentary assessments (e.g., self-report multiple times a day); passive collection via technology (e.g., Fitbit; texting behaviors); or psychophysiological data (e.g., fMRI; heart rate). Data requirements are at least 60 time points (more is better!) and at least 5 variables (up to 20).

  • Intermediate to advanced

  • It would be helpful to have some familiarity with R and to be comfortable with statistical testing and inferences.

  • R

  • Participants will be provided with a curated reading list and article pdfs along with recorded videos, slides, code, and toy data.