| Module 3: Interpreting Data
7.2 Case Study: Lung cancer and smoking
Another much discussed issue has been the relationship between smoking and lung cancer.
Let us say that you begin by noting that more people seem to be dying of lung cancer. In the past it had been a relatively rare disease. You start gathering information about a group of people who died of lung cancer. You want to know if something is causing this increase in lung cancer. You are looking for an explanatory variable. You are asking, "Is there something that this group is doing differently from the general population that is causing them to have a higher incidence of lung cancer than the general population?” This process is shown in Figure 7.2.1
Figure 7.2.1
In an experiment, levels of an explanatory variable are controlled by the researcher in order to establish the effect on the response variable. In a retrospective study , the researcher is attempting to identify the explanatory variable after the event. In this example, the explanatory variable is thought to be the level of smoking.
Retrospective studies are based on a process called induction, illustrated in Figure 7.2.2. In these studies, you hope to establish a general statement by referring to a lot of examples.
Induction
Figure 7.2.2
The deductive process, outlined in Figure 7.2.3 for your information, is different form of investigation.
Deduction
Figure 7.2.3
How could we attempt to establish strong association so that the evidence of a link between an explanatory variable, X, and a response variable, Y, is as strong as it can be?
If the association between X and Y is measured in many different situations then this suggests that the relationship is not due to confounding variables. For example, to provide evidence that smoking causes lung cancer then it is important that studies are conducted over a range of situations such as different countries, environments, with different units and over long periods of time.
There exists a plausible explanation that shows how X could cause the observed changes in Y. For a long time the tobacco industry argued that the association between smoking and lung cancer could be due to a common response. A genetic factor was put forward as a possible common response by R.A Fisher (1974), illustrated in Figure 7.2.4.
Figure 7.2.4
How plausible is this argument? Such a hypothesis cannot explain the different patterns in the rise in lung cancer that have been observed within both sexes. For example, studies indicate that the risk of death from lung cancer was related to the number of cigarettes that a person smoked and the age at which he/she began smoking. Also, the death rate from lung cancer in males was directly related to the prevalence of smoking with a 30-year time lag and as the prevalence of smoking in women increased a corresponding increase in lung cancer was observed.
No equally plausible third factor exists that could cause the changes to X and Y together. The existence of a third factor was not supported by the data collected from the studies that were conducted.
You need to examine critically any causal claims based on observational data. As we know from earlier modules observational data is often the only form of data available. TAKE CARE!
|  |
