Testing for goodness rather than lack of fit of continuous probability distributions