Let's think of the case of one hospital. There are 400 patient of heart attack in the hospital. After one month, 72 patients is dead and 328 patients are alive. The question is could we estimate the mortality rate? In here, we can check frequentist paradigm first.
Frequentist paradigm
Before dealing with the mortality rate, we need to establish our referece population. The reference population can be heart attack patients in the region or the number of heart attack patients in the hopistal long time. Both model can be reasonable. But the problem is our 400 patient data is not random sample from either of those population model. If we want to deal with random situation, we need to consider every people who can go hospital and have potential to heart attack. As you can see, This is hypothetical case and there are philosophical issues
Bayesian paradigm
We can start from setting reasonable distribution. In here, there are only two choices heart attack or not. So, we can use binomial distribution(also bernouilli distribution). Then, the distribution and probability can be written liek this. In the equation, the y tilde is meaning vector from probability density function for entire set. Yi∼B(θ)
Likelihood function
Now, we get probability function. We can go further with this equation to get the mortality value. The likelihood function is density function thought as a function of theta like this. L(θ|y)=Πni=1θyi(1−θ)1−yi
Through this method, we can calculate the parameter value which maximizing likelihood. This method can be applied to another distribution also like exponential distribution.