for the hazard rate (or ``conditional probability of failure'') in a time interval (t_i,t_{i+1}). Here d_i is the number of deaths in (t_i,t_{i+1}), c_i is the number of censored observations, n_i is the number at risk just before time t_i, and D_i=t_{i+1}-t_i is the length of the time interval so that h_i is a rate and not a probability. The expression   n_i - (1/2)c_i   corrects for a variable number of censored observation in (t_i,t_{i+1}) but not for a variable number of deaths. This handout assumes that you know something about maximum likelihood estimators.)