Skip to content

Growth Curve Analysis Mixed and Fixed Effect Models

Jump to bottom
danibachar edited this page Aug 3, 2020 · 3 revisions

Growth Curve Analysis

As both R and Python are scripting/programming languages that are being used for statistics, and GCA is kind of Machine Learning statistic tool we are going to go over some basic concepts and tutorials to understand the ideas and heuristics behind the data pre-processing, and analysis. In addition, we try and unveil the ideas and pinpoints on how to build a good predictive model.

In the Psychology field, and Eye Tracking analysis Growth Curve was used in many types of research. We recommend reading Daniel Mirman Book and Website and especially the last chapter where he brings concrete examples on how to build Mixed Effect Models. "Growth curve analysis (GCA) is a multilevel regression technique designed for analysis of time course or longitudinal data. A major advantage of this approach is that it can be used to simultaneously analyze both group-level effects (e.g., experimental manipulations) and individual-level effects (i.e., individual differences). We have been using this method for several years, particularly in the context of visual world paradigm (VWP) eye tracking data and learning curves, though it can be applied to any time course data."

Statistical Models

In general, most of the notebooks we are going to work with, from this repository are for data pre-processing. As you probably know, we would like to process the data, clean it, and alter it before running the model. We are going to focus on 2 main Machine Learning learning models for predicting and analyzing Growth Curves.

  • Linear Mixed-Effects Models
  • Generalized Mixed-Effects Models

Time Bins

Growth curve analyses model changes over time, to conduct a Growth Curve Analyses for a visual world study, the data for every trial must be aligned at a particular time point in the sentence (e.g., verb offset).

Time is measured in small windows relative to that synchronization point (ranging from 17 ms to200 ms) In most trials we have encountered so far, the time windows are measured in 20ms timestamps, this gives us a good approximation to a video frame rate (60FPS). But one can think of how the time window can affect his/hers study and try to change it.

The measure of interest is whether the participant is looking at a particular object (the matching picture) during each of these time windows. Trials are averaged together within a participant to get a proportion of looks to a given item during that window.

Model Variables

  1. In general

Data distribution and underlying assumptions

  1. Normal Distribution of relevant data of variables

Links

Clone this wiki locally