A binomial experiment consists of $n$ trials. If we want to define the probability of success as $p$, which assumption is NOT required?