Statistics for Business: Decision Making and Analysis, 3rd Edition

... because the residuals zoom in on the deviations from the fitted line. The analogous plot of e on $\widehat{Y}$ is useful in multiple regression. The scatterplot in Figure 23.5 graphs the residuals on the fitted values for the multiple regression of sales per square foot on median income and number of competitors.

From this plot, we can visually estimate that s_e is less than $75 per square foot; all of the residuals lie within $150 of the horizontal line at zero, the average of the residuals. (The Empirical Rule suggests about 95% of the residuals are within ±2s_e of zero.) In fact, s_e = $68.03. As in simple regression, the residuals should resemble a swarm of bees buzzing around the horizontal line at zero, with no evident pattern.

The most common uses of the scatterplot ...