Answer: The second option is the only good residual plot for a good linear fit.
The last graph may also be related to a linear relation, but it is not a good fit (so the actual slope of the relationship is smaller than the slope of the fit, this is because at the beginning the points are above the line and at the end, the points are below the line, this may be related to a bad measure)
Step-by-step explanation:
Ok, it depends on the origin of the residual plots, the common case will be for a linear fit, so:
If the residual plots are of a linear fit, then we need to see the points near to the line and with a random distribution.
The only option that meets these two things is the second option. (in all the other options we can see a lot of variances or a pattern, like in the first or in the last)
The third is kinda random and may be related to a linear relationship that was measured very poorly.
In the last graph we can see a pattern, this may mean that the actual slope of the line is different that the slope of the linear fitt, so this relation is linear but not well fitted, which may be because of a bad measure.
