Scribd
Upload
3K views

Bayesian Notes

Some notes comparing Bayesian and Frequentist Inference

Uploaded by

BirthDeath
Copyright
© Attribution Non-Commercial (BY-NC)
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
3K views

Bayesian Notes

Some notes comparing Bayesian and Frequentist Inference

Uploaded by

BirthDeath
Copyright
© Attribution Non-Commercial (BY-NC)
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
You are on page 1/ 2
Frequentist Approach and Bayesian Approach + Yo)- Joint distribution p(y(6) = TT yy; from p(y,(9), independently and identically distributed, i= 1, 2, ....n. (y, |@), as a function of y. Likelihood given the data y = p(y/0), as a function of 8. Frequentist Approach Bayesian Approach Joint distribution of y = p(yi®) ‘Sample summary (Statistic) T(y) related to the inference problem of interest. Examine the sampling distribution of T(y) (ce, the collection of all possible values of ‘T(y) when y varies in the sample space) How to obtain this sampling distribution? Find the probability that T(y) is in a certain subset in the sample space using the sampling distribution of T(y). Cried val Use a statistical method based on T(y) for the inference problem of interest. Make a decision based on the method's repeated sampling performance (long run sampling performance). Expectation (mean) of T(y) w.r-t. y Variance of T(y) w.t.t. ¥ Vis «-+ Yn from N(Q,0?) with known o Sampling distribution of the sample mean ¥ is N@,o"/n) 95% confidence interval for @: (¥ -1.96fo/Vn], F + 1.96[o/Vn)) Meaning in repeated sampling? Prior distribution of © = p(6) Joint distribution of “@ and y” = p(®)p(yl9) Posterior distribution of @ given the data y= Ply) < pO)p(vi6) How to obtain this posterior distribution? Calculate the probability that @ is in a certain subset of the parameter space using the posterior distribution. Use the posterior distribution to solve the inference problem of interest. Make a decision based on the posterior probability (and a relevant utility function). Expectation of @ w.r.t.@ from the posterior p(Oly) Variance of @ w.r. 6 from the posterior p(Gly) | Yis vs Yo fom N(@,o*) with known Posterior distribution with a uniform prior on 0 is N(¥,0"/n) (will be different for different priors) 95% credible interval for 0 (¥ -1.96[o/Vn], ¥ + 1.96[o-/n}) Meaning in posterior probability? TT ‘Advantages of Bayesian Methods 1. Interpretation Having a distribution for your unknown @ is easier to understand than a point estimate and a \Y standard error. In the Bayesian framework, we can make statements concerning the probability Of 6 falling in an interval, by specifying a prior distribution for 0. Thus, additional structure (i.e. a prior for 0) leads to an ease in interpretation. 2. Bayesian inference obeys the likelihood principle 3. Bayesian inference does not lead to absurd results “Absurd results can be obtained when doing UMVUE estimation. Suppose y~ Poisson (2), and sve want to estimate 9 = e—, 0<0

You might also like