Correlation & Causation
Before moving on, let's take a careful look at what we are talking about. Here are the definitions of correlation and causation:
Since we know the definitions, we can ask the main question.
What is the relationship between correlation and causation?
Here are a series of diagrams to help you understand the difference between the two.
- Correlation: the relationship between things that happen or change together.
- Causation (=causality): the relationship between somethings that happens or exists and the thing that causes it.
Since we know the definitions, we can ask the main question.
What is the relationship between correlation and causation?
Here are a series of diagrams to help you understand the difference between the two.
The diagram above shows three factors that we are interested in; factor A, factor B, and factor C. Since we are mainly going to talk about the relationship between A and B, we will begin by saying that there is a degree of uncertainty involved in the existence of C. This is shown by the dotted circle around C. We will return to factor C later.
Before any measurement or observation, we have no idea whether there is a clear relationship between A and B. Thus, there exists some uncertainty between A and B, which is shown by the dotted line. Either A and B can have some kind of definite relationship or they can be entirely unrelated to each other.
This is where the notion of correlation comes in. If there exists a consistent or systematic relationship between A and B, then one can say that there is a correlation between A and B. In this case, A and B can exhibit various relationships. For instance, when A and B are measurable quantities, B may increase while A decreases. The relationship can be of either quantitative or qualitative character, and correlation is detected via measurement (quantitative) or observation (qualitative).
So one might finally ask the question; how is this different from causation?
Well, here is the difference.
So one might finally ask the question; how is this different from causation?
Well, here is the difference.
Let's go back to the beginning and say that factor C also exists in our scheme. Now our question concerns how each factor affects the other. In the following diagrams, we will represent causation with vectors (or arrows). The example above shows that A causes B, B causes C, and C causes A. The inverse of the relationships may not be true unless proven by further studies.
With our third diagram, we have established that there is some identifiable relationship (a correlation) between A and B. However, we cannot conclude that the identified relationship is a casual one. This is because there are several possible 'interactions' between the three factors that contribute to the correlation. Here are the possibilities:
Case 1: The correlation is actually a causation
Case 2: The correlation is due to relationships between 'A and C' or 'B and C' (nothing causal between A and B)
Case 3: There are other factors affecting the correlation
The example above clearly shows why it's hasty to say that correlation implies causation. There are multiple combinations of relations that can lead to the observed correlation. Furthermore, there can be other factors that contribute to the correlation. Thus, causation can only be determined through controlled experiments and systematic accumulation of experiment results.
Concluding Remarks
The demonstration above shows the difference between correlation and causation. At the same time, it shows how it can be very difficult to determine whether a correlation indicates a causal relation. Unfortunately, knowing the difference alone is not enough to determine causality between two factors. The 'Nature of statistics' section and the 'Human Nature' section will further explore the difficulties involved in effectively evaluating and using statistics and similar information.
Here is also an interesting, in-depth article on how one can work on correlation and causation:
Here is also an interesting, in-depth article on how one can work on correlation and causation: