[ad_1]
Why does correlation not equal causation for time series?
In time series analysis, it is important to understand whether one series affects the other. For example, it is useful for commodity traders to know whether there has been an increase in item A leads to an increase in item B, However, originally, this relationship was measured in the 1980s using linear regression clive granger And Paul Newbold showed that this approach gives incorrect results, especially for non stable time series. As a result, he conceived the concept of cointegration, which won Granger the Nobel Prize. In this post, I want to discuss the need and application of co-integration and why it is an important concept that data scientists should understand.
Overview
Before we discuss co-integration, let us discuss its necessity. Historically, statisticians and economists used linear regression To determine the relationship between different time series. However, Granger and Newbold showed that this approach is wrong and leads to something called spurious correlation,
A spurious correlation is one where two time series may appear to be correlated but in fact lack a causal relationship. It’s a classic’correlation does not mean causation‘ statement. It’s dangerous because even statistical tests can say that it’s simple relationship,
Example
An example of a spurious relationship is shown in the plot below:
Here we have two time series But) And B









