Analysis of Residuals
Histogram of Residuals
About this App
This app is supposed to help users practicing how to chose the best performing linear model for a given dataset. It supplements my article on how to select the best performing linear regression for univariate models
Although there are various performance indictors, I personally recommend to have a look at two factors:
This indicator illustrates how much variation is explained by your model. In contrast to the simple R2, the adjusted R2 takes the number of input factors into account. It penalises too many input factors and favours parsimonious models.
The residuals are the difference between your predicted values and the actual values. Their benefit is that they can show you both the magnitude as well as the direction of your errors.
With the residual plot you can determine if your model has areas with an upward- or downward bias. Neither of them should occur systematically. So make sure your residuals are equally distributed around zero.
Finally, the residual's histrogram shows the bandwidth of your errors and illustrates how certain your predictions are.