For the purpose of analyses, they should be independent, identically distributed random variables. But the ideal is that they are all 0.

An analysis of the residuals can be used to check that the modelling assumptions are appropriate.

the sum of the residuals is always 0(zero) because its how far they are away from the LSRL. that's what makes them residuals in the first place :P

how much in residuals is jason alexander receiving from seinfeld

One limitation of statistics is that everyone doesnÕt understand them or how to use them. There is also a possibility for errors.