What does difference in the number of dots in Fitted vs Residuals plot mean?

Like in the below plots of two different models:

Model 1: enter image description here

Model 2: enter image description here

The length of fitted() of both models is the same, but why does it seem that Model 2 has a lot more predictions?

Also does this kind of difference tell anything about the goodness of fit? Is more dots necessarily better?

The models are:

model1 <- glm(formula = cancer ~ exposure + skin + gender, family = binomial, 
    data = dta)
model2 <- glm(formula = cancer ~ exposure + age * trt * skin + gender, 
    family = binomial, data = dta)

1 Answer 1


There is probably some overplotting going on in the plot which appears to have too few dots. Try plotting the plot "by hand" and then add some jitter() and you will be able to visually confirm that.

  • $\begingroup$ Is overplotting bad or does it just mean that there are many predictions getting the same predicted value and the same residual? $\endgroup$
    – mavavilj
    Commented Oct 23, 2016 at 12:30
  • $\begingroup$ The latter. It is only a problem for the visual investigations of the plot. I guess it is in the nature of your data, that only limited numbers occur and thus equal predictions and residuals are frequent in model1. It is not a problem per we. $\endgroup$
    – Bernhard
    Commented Oct 23, 2016 at 13:19

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.