Difference between confounding and interaction What is the basic difference between confounding and interaction? Is it possible to occur both at the same time in data? Can anyone please explain this plainly and with an example? 
 A: A confounding variable is a variable that correlates with both your regressor and the dependent variable. In some way, this second predictor variable explains all or part of the dependent variable and also is reflected in the independent variable. In essence they share a common quality that means when both are included that quality is over-represented.
In an ecological system, something like a disease that kills both predator and prey is acting on the populations of both, yet has nothing to do with the effect of predation on the decline of prey or growth of predators. It confounds the true predator-prey relationship, particularly if it is disproportionate in its virulence between species.
Interaction is much more complicated because it means that two separate regressors work together to create an outcome variable. They do not overlap, they in some way coalesce in an effect that is not simply additive. Their relationship, as it acts on your dependent variable,  is sometimes difficult to figure out.
If you have a situation where two proteins work together to accomplish some kind of chemical process in the human body with only one pathway. Removing one or the other will break your model. Though it may be difficult from the model to exactly quantify their relationship if there are other components which create the appropriate environment for the reaction or regulate the presence of the resulting product (like reuptake or conversion).
With confounding variables, you can often leave one or the other out and get a more accurate model (although not always). With an interaction, leaving one or the other out will likely make it worse.
