Answer: Compares the value of the predicted value and the actual value
Is random when the function fits the data well
Explanation:
The residual data is a comparison between the predicted or fitted data with the actual measured data. So Compares the -value of the predicted value and the actual value is correct.
You always want to see randomness, because if you see a pattern in the residua means that the model does not fit well the data, so "Is random when the function fits the data well is also correct."
If most of the points lay upside the predicted value, we will have a positive-sum. It is common to see it for quadratic functions, but it does not happen always, so the last sentence is not true.