This repository contains the full dataset and the scripts for generating the analysis and the figures from "Dissociating linear and angular path integration in ageing and Alzheimer’s disease"1
The anonymised dataset for healthy elderly and MCI patients is the one from the "Differentiation of mild cognitive impairment using an entorhinal cortex-based test of virtual reality navigation"2
Clone the repository anywhere in your machine using git clone.
Open the folder in Matlab by making sure to add folders and subfolders to the path.
Developed using Matlab R2021b.
Requires the following MATLAB toolboxes:
- Statistics and machine learning
- Optimization toolbox
- Global Optimization toolbox
- Ecoometric toolbox
- Mapping toolbox
We use a customised version of Rob Campbell (2023) sigstar which is included already in the project files.
We use a fast version of the DeLong algorithm (Sun and Xu, 2014) whcih is already included in the project files.
Scripts to reproduce the plots and analysis in the paper can be found into the Main folder.
Each of the scripts is standalone, meaning that will perform the entire stages of the analysis (i.e. data loading, pre-processing, behavioural analyisis, model fitting, plotting) and it will produce one or more plots from any given figure of the main text/supplementary as specified in the script name.
Plots will not be displayed within Matlab but output will be saved inside Output folder.
Please be aware that some scripts will take longer times than others to complete. If you want to run the entire analysis, you can do it with RunAllAnalysis.
Below you can find some more information about the analysis process with some references to relevant scripts/functions.
To allow the model to be run with different configurations we have used a config struct variable. To see information about properties set in the struct please refer to GLAMPI_PrepareBaseConfig. To run a specific version of our generative model the variables ModelName,ParamName and NumParams have to be correctly set. Please see Fig_2_S3_S4_ModelSelection.m script to see all available combinations of the model.
Details for the preprocessing of the data is described in the online methods. The script responsible is GLAMPI_PreprocessData. Other important functions are TransformPaths aligning the paths, CalculateTrackingPath for using tracking data to extract information about the participant behaviour while performing the outbound path and CalculateOoB to recover relevant information from the trials marked as out of boundary.
The entry point for calculating behavioural performances and fitting the model is GLAMPI. In PerformGroupFit behavioural data is calculated. FitData executes the model fitting using the specified model (see config struct variable) with the limits for each of the fitted parameters. The fitter use fmincon Matlab function, but adopts a global search approach to avoid the local minima problem.
For visualizing trials you can use the VisualizeRealtimeTrackingData function.