f its data is categorical, a population is called multinomial. That belongs to a non-overlapping classes discrete collection. Here fi( observed frequency) is equal to an ei(expected count) which is the null hypothesis for goodness of fit test for multinomial distribution. so fi=ei During each trial, multinomial experiment is exactly the same as a binomial experiment. Probability that a particular outcome will occur on any given trial is constant. Outcome on one trial does not affect the outcome on another trial when the trials are independent