The descriptive function of epidemiology is important. But the greatest contributions of epidemiologic science have come from its other main function, causal inference. This is a brief introduction to the concept of causal inference in epidemiology, which will inform some of our discussions throughout the course. As a species, we have a tendency to explain the world around us in causal terms. We instinctively compare the outcomes of our actions with outcomes that followed different actions, and if the two outcomes differ, we tend to explain these as a causal effect of the actions on the outcome. This way of thinking has developed over thousands of years of our evolution, and it works quite well in daily life. However, this approach is not as effective in research and can lead to serious misinterpretations. In epidemiological research, we typically compare two population with each other, with regard to exposure and outcome. We call exposure, any potential causal characteristic such as behaviors, environmental factors, treatments, occupation, genetic factors, and so on. The outcome is most often a disease. Although, it can be any other condition of interest. The fact that the populations were compared different, limits our ability to speak about cause and effect. In theory, the only way to be certain about the causal relationship would be the following, we observe a population exposed to a certain factor, and report the outcome of this exposure. For example, the proportion of the population that developed the disease. Then, we get into a time machine and travel back in time. This time, we observe the exact same individuals without exposing them to the exposure under study, and report the outcome. If the two outcomes differ, the only possible explanation is that there is a causal relationship between the exposure and the outcome. Unfortunately, time travel is still not among the tools we have in epidemiology. So, such experiments are impossible. We have to stick to comparing different populations, which essentially forces us to study only associations between exposures and outcomes. Does this imply that we can never prove causality through epidemiological research? Yes and no. Considering that we can never study the same population twice under the exact same conditions, there's no method to produce irrefutable evidence of causality. On the other hand, if the two population groups that we used for comparison are very similar to each other or exchangeable, as we call them, we can be quite confident that the association we detect is of causal nature. This leads me to the main point I would like to highlight. The statistical analysis of data generated by epidemiological studies can only provide you with evidence that, an association between the exposure and the outcome exists. It is up to you then, to decide whether it is reasonable to take the extra mental step and declare with little or much confidence that the exposure is what causes the outcome. To sum it up in a sentence, you should always keep in mind that association does not necessarily imply causation. Epidemiological knowledge is essential to decide when association implies causation.