Multivariable Time Series — Approach Guide for Time Series with Multiple Predictors
I. The Downsides of Univariate Time Series Models
Traditional time series models, like autoregressive integrated moving average (ARIMA), rely on past values of the target variable to make predictions. While this approach works well for univariate time series data, it falls short in scenarios where the target variable depends on numerous predictors. Real-life situations often involve complex relationships among variables that go beyond simple lagged dependencies.
In this guide, I will walk you through the appropriate approach for this problem and we’ll explore the nuances of multivariable time series modeling and when to choose it over linear regression.
II. Solutions: Multiple Linear Regression vs Multivariate Time Series
In this session, I will quickly go through the benefits of each model, then list down some criteria when to choose one and also give you some examples to illustrate the concept.
1. Linear Regression:
- Simple Linear Relationships: When the relationships between predictor variables and the target time series are primarily linear and do not involve complex temporal…
