What is a posterior distribution?

What is a posterior distribution?

• A. A distribution of posterior probabilities reflecting beliefs after considering the data.
• B. A distribution of sampling errors before data collection.
• C. A distribution of observed frequencies in the sample.
• D. A distribution used to compute p-values.

Answer: (A)

The posterior distribution is the updated view about a parameter after observing data. It’s a probability distribution over the possible values of the parameter, representing our beliefs new after taking the data into account. In Bayesian reasoning, we start with a prior belief about the parameter, then combine that with how likely the observed data are under each possible parameter value (the likelihood). The result is the posterior, which tells us how plausible each parameter value is given both what we believed before and what we actually observed. Think of estimating a coin’s chance of landing heads. If you start with a prior belief about that chance (before any flips) and then flip the coin many times, the posterior distribution updates your belief to reflect the data. If you use a Beta prior and observe k heads in n flips, the posterior becomes Beta with parameters updated by the data, giving you a full distribution over the possible probabilities of heads. This posterior then lets you summarize uncertainty with a mean, credible interval, etc. It’s not about the distribution of sampling errors before data collection, nor the observed frequencies in the sample, nor a distribution used to compute p-values. It’s specifically the distribution of parameter values after updating with the data, representing our updated beliefs.

The posterior distribution is the updated view about a parameter after observing data. It’s a probability distribution over the possible values of the parameter, representing our beliefs new after taking the data into account. In Bayesian reasoning, we start with a prior belief about the parameter, then combine that with how likely the observed data are under each possible parameter value (the likelihood). The result is the posterior, which tells us how plausible each parameter value is given both what we believed before and what we actually observed.

Think of estimating a coin’s chance of landing heads. If you start with a prior belief about that chance (before any flips) and then flip the coin many times, the posterior distribution updates your belief to reflect the data. If you use a Beta prior and observe k heads in n flips, the posterior becomes Beta with parameters updated by the data, giving you a full distribution over the possible probabilities of heads. This posterior then lets you summarize uncertainty with a mean, credible interval, etc.

It’s not about the distribution of sampling errors before data collection, nor the observed frequencies in the sample, nor a distribution used to compute p-values. It’s specifically the distribution of parameter values after updating with the data, representing our updated beliefs.