Selling price of houses in two years
This diagram again shows that the marginal relationship between Y and X may be misleading if the data are observational.
The reason is that a larger proportion of sales were in poorer districts in 2004.
The marginal relationship was caused by a lurking variable, District.
The data are artificial.