On using non identical and independent Bernoulli trials in human surveys