"In conclusion we processed data from a group of people that contained smoking status, living status at the time of the second point and age at the second point of study. \n",
"We have found an inverse association between smoking status and mortality at first glance. The fact of being a smoker was associated with a smaller mortality rate. This could make us believe that smoking protects us from death. However, to evaluate the true and direct effect of smoking on health status we should take into consideration the variable age because age affects your chances of death and smoking affects the age at which people where taken into consideration. Indeed we could end up with a population where elderly people are smokers and die less frequently because all the people susceptible to adverse effect of smoking have already died (for exemple of lung cancer). This would constitute an important bias.\n",
"The bias \n",
"To sum up the Simpson's paradox we analyzed comes from the fact that there is an association between two variables X (Smoker status) and Y (mortality rate), that is inverted when we condition on a third parameter (Age class). This is very clear if we took into consideration the causal DAG connecting the three variables. "
