{"id":75191,"date":"2024-12-12T02:13:18","date_gmt":"2024-12-12T07:13:18","guid":{"rendered":"https:\/\/world.edu\/?p=75191"},"modified":"2024-12-12T02:13:18","modified_gmt":"2024-12-12T07:13:18","slug":"if-correlation-doesnt-imply-causation-how-do-scientists-figure-out-why-things-happen","status":"publish","type":"post","link":"https:\/\/world.edu\/if-correlation-doesnt-imply-causation-how-do-scientists-figure-out-why-things-happen\/","title":{"rendered":"If \u2018correlation doesn\u2019t imply causation\u2019, how do scientists figure out why things happen?"},"content":{"rendered":"

\"\"<\/p>\n

Most of us have heard the phrase \u201ccorrelation does not equal causation\u201d. But understanding how scientists move beyond identifying correlations to establish causation remains a mystery to many.<\/p>\n

Finding out what causes a particular outcome is often the primary goal of scientific research, especially in studies relating to our health.<\/p>\n

We want to know if a certain factor \u2013 say, drinking wine or eating chocolate \u2013 will lead to better or worse health outcomes. That way, we can make more informed decisions about our health.<\/p>\n

But how do scientists actually get those answers?<\/p>\n

Correlation versus causation: the basics<\/strong><\/h4>\n

It\u2019s easy to find examples of correlations where two variables are linked, but there\u2019s no causal relationship. For instance,\u00a0there\u2019s a correlation between chocolate consumption<\/a>\u00a0and the number of Nobel Prize winners per capita in a bunch of countries.<\/p>\n

Does eating chocolate cause people to win Nobel Prizes?\u00a0Of course not<\/a>.<\/p>\n

\"\"

Correlation between countries\u2019 annual per capita chocolate consumption and the number of Nobel laureates per 10 million population. The New England Journal of Medicine \u00a92012<\/p><\/div>\n

This correlation likely exists because chocolate consumption serves as a proxy for wealth. In turn, wealth relates to educational opportunities and funding for completing high-quality research that might lead to a Nobel Prize.<\/p>\n

It\u2019s not enough to just find a link between two things. Scientists need much more evidence before we can start to assume a causal relationship.<\/p>\n

Building a case<\/strong><\/h4>\n

In chemistry or physics, it\u2019s often possible to conduct experiments under highly controlled conditions to understand how X affects Y. When it comes to human biology, it\u2019s rarely so simple.<\/p>\n

In most instances, to establish causality we use indirect evidence (more on that in a moment). It requires an approach called inductive reasoning \u2013 a process where scientists make generalisations\u00a0based on the available evidence<\/a>.<\/p>\n

It\u2019s a bit like how a prosecutor might build a criminal case based on circumstantial evidence. While a single piece of such evidence might not be persuasive on its own, as the pieces add up, they strengthen the case.<\/p>\n

There\u2019s one interesting contrast, however. In criminal cases, the stakes are incredibly high, and the threshold for proof is \u201cbeyond reasonable doubt\u201d. In science, when we make the case for a causal relationship, it\u2019s usually based \u201con the balance of probabilities\u201d.<\/p>\n

This lower threshold of proof reflects the fact scientists are happy to revise their beliefs if and when better evidence becomes available.<\/p>\n

\"\"

Inductive reasoning starts with observations. The Conversation<\/p><\/div>\n

Types of indirect evidence of causation<\/strong><\/h4>\n

The type of indirect evidence scientists use to infer causation\u00a0can take different forms<\/a>. These include:<\/p>\n

1. Temporality<\/strong><\/p>\n

This is the only absolute requirement for a relationship to be causal. That is, an exposure must occur\u00a0before<\/em>\u00a0the outcome for an exposure to\u00a0cause<\/em>\u00a0an outcome.<\/p>\n

As obvious as this appears, there can be situations where this isn\u2019t clear cut. For example, there might be a long lag time between the two events. For example,\u00a020\u201360 years<\/a>\u00a0can pass between exposure to asbestos fibres and development of mesothelioma, a type of cancer.<\/p>\n

Or it might not be immediately obvious what is the exposure and what is the outcome: do sleep disorders lead to depression, or is disordered sleep a symptom of depression?<\/p>\n

2. Strength of association<\/strong><\/p>\n

A\u00a0strong<\/em>\u00a0association between two variables is generally considered suggestive of a causal relationship. That is, if one thing happening means another thing is likely to occur, we generally consider this good evidence for causality.<\/p>\n

For example, studies showing that high consumption of alcohol is associated with liver damage demonstrate a strong effect. Therefore, they\u2019re highly supportive of a causal relationship.<\/p>\n

3. Consistency across studies<\/strong><\/p>\n

If various studies using different methods all yield the same or similar associations, this also supports the existence of a causal relationship.<\/p>\n

We generally have more confidence in scientific findings when they can be replicated using different study approaches.<\/p>\n

4. A plausible mechanism exists<\/strong><\/p>\n

Being able to demonstrate a mechanism that could explain the association between an exposure and outcome provides further support for a causal relationship.<\/p>\n

For example, if lab or animal studies show how a substance damages cells, this would be supportive of a causal relationship between this substance and disease in people.<\/p>\n

5. Dose-response relationship<\/strong><\/p>\n

Observing that higher exposures lead to stronger effects is considered highly supportive evidence for a causal relationship.<\/p>\n

However, it\u2019s important to note sometimes there\u2019s a threshold effect when it comes to causation. That is, an exposure doesn\u2019t cause disease until it reaches a particular level. This is generally true for infectious diseases, where a minimum infectious dose is required before a person is likely to get ill.<\/p>\n

The power of randomised controlled trials<\/strong><\/h4>\n

Indirect evidence usually plays an important role in inferring causality. But there\u2019s one type of study that\u2019s the gold standard for\u00a0providing direct evidence of a causal relationship<\/a>. It\u2019s called a randomised controlled trial, or RCT.<\/p>\n

In an RCT, participants are randomly assigned to either receive an intervention or to be a \u201ccontrol\u201d. This ensures if you see a difference between the two groups, this can only be due to the effect of the intervention. It effectively proves there\u2019s a causal relationship.<\/p>\n

You can think of it as the equivalent of catching a criminal red-handed. Unfortunately, due to ethical and practical considerations, we don\u2019t always have evidence from well-conducted RCTs.<\/p>\n

For example, we don\u2019t have RCT evidence that smoking causes lung cancer. The reason is that the strength of the indirect evidence supporting a causal relationship is so compelling, it would be unethical to do these studies.<\/p>\n