"Big data" problem

We use many statistical and mathematical processes to analyze data in order to find the correlation between some certain variables we want. However, the problem of such process always exists. We can always omit some important but unobvious variables. Moreover, even when we find there is a strong correlation, we may still be uncertain about the causality between the variables. Of course, we can always improve our methodology and reduce the impacts of all sorts of errors. However, firms which analyze data to serve their interests need to put their models in a game theory model as all parties tend to run similar analysis.