# GLACIAL Paper Review

*By Aaditya Panchal*

## Introduction & Context

**GLACIAL** stands for "Granger and LeArning-based CausalIty Analysis for Longitudinal studies." It's designed to discover causal relationships in data where multiple individuals are observed over time but sparsely sampled. This is typical in longitudinal studies, especially in fields like population health.

**Granger Causality** is depicted as a statistical method to determine correlation between two sets of data in a timeseries.

This can be useful in identifying relationships between two(or more) independant variables. In terms of machine learning, we can look at the similiarities or differences between two studies, or the stages in which a disease progressed in two different people.

Although it is only succeptible under the conditions:

* Densely sampled time series: data collected at equal intervals over a period of time.
* Time-varying signals: any quantity or set of data points that changes over time. Their values are dependent on the specific point in time at which they are observed.

## Why GLACIAL?

GLACIAL was developed in order to handle complexities such as individual specific dynamics and missing data. This is because traditional granger causality struggles in longitudinal analysis due to sparse and irregular sampling, missing data, etc.

* It combines traditional Granger causality principles with machine learning, using a neural network to learn dynamic, non-linear relationships and handle missing data.
* GLACIAL uses a train-test method to see if it can accurately predict the outcomes of some people's data. Firat, it leaves aside some data, in the training phase. After it is trained with the rest of the data, it tests itself, on the left out data, to see if it can predict the causal relationships in what happens.