With this approach I would need a separate loop for each response variable. I’m making pretty fundamental graphs since these are exploratory plots, not publication-prepared plots. I will make a scatterplot and add regionally weighted regression lines by way of geom_smooth(). I use such lines with nice warning, as it may be straightforward to get too attached any sample the loess line shows.
Sometimes there’s really nothing wrong with your mannequin. In the above example, it’s fairly clear that this isn’t an excellent model, however typically the residual plot is unbalanced and the mannequin is kind of good. Now should you’d collected information every single day for a variable known as “Number of energetic lemonade stands,” you could add that variable to your model and this problem can be mounted. That is, there’s fairly a few datapoints on both sides of zero that have residuals of 10 or larger, which …