The difference between the predicted value (often from a regression equation) and the actual or observed value is termed the residual value. Residuals reflect the overall badness of fit of the model. Examination of residuals in regression analysis will identify atypical cases. Ideally, the residuals should have constant variance along the line. A normal probability plot of the residuals can check this. In the plot of residuals against the explanatory variable (or the fitted values), there should not be any pattern if the assumption of constant variation is met, i. e. residuals do not tend to get larger as the variable values get larger or smaller.