Omitted Variable Bias
The difference between the value of an estimated parameter and its true value due to failure to control for a relevant explanatory (confounding) variable or variables. It is often possible to assess the direction of the bias by using common sense. For example, if a regression of hospital costs finds that the cost per patient episode is higher in teaching hospitals than in non-teaching hospitals, the inference that teaching hospitals are less cost-effective than non-teaching hospitals is likely to be false because their costs are in reality increased by the presence of teaching - a variable that was omitted. So the bias is clear. It is probably better on the whole to err on the side of including the wrong variables than to omit the right ones. Trends over time are a potent source of omitted variable bias. Consider the following figure. Kocj discovered the tubercle bacillus in 1882 but effective treatment for res piratory tuberculosis began only in 1947. The fall in the death rate from this disease is manifest.
But seen against a longer time frame in the second figure, the impact of treatment appears much less dramatic. Plainly something else was going on besides the new treatment. It seems that curative medical measures played little role in mortality decline. What these factors may have been is argued about.