Lesson 6.3 Posterior predictive distribution

Loading...

Del curso dictado por University of California, Santa Cruz

Bayesian Statistics: From Concept to Data Analysis

1265 calificaciones

University of California, Santa Cruz

Bayesian Statistics: From Concept to Data Analysis

1265 calificaciones

This course introduces the Bayesian approach to statistics, starting with the concept of probability and moving to the analysis of data. We will learn about the philosophy of the Bayesian approach as well as how to implement it for common types of data. We will compare the Bayesian approach to the more commonly-taught Frequentist approach, and see some of the benefits of the Bayesian approach. In particular, the Bayesian approach allows for better accounting of uncertainty, results that have more intuitive and interpretable meaning, and more explicit statements of assumptions. This course combines lecture videos, computer demonstrations, readings, exercises, and discussion boards to create an active learning experience. For computing, you have the choice of using Microsoft Excel or the open-source, freely available statistical package R, with equivalent content for both options. The lectures provide some of the basic mathematical development as well as explanations of philosophy and interpretation. Completion of this course will give you an understanding of the concepts of the Bayesian approach, understanding the key differences between Bayesian and Frequentist approaches, and the ability to do basic data analyses.

De la lección

Priors and Models for Discrete Data

In this module, you will learn methods for selecting prior distributions and building models for discrete data. Lesson 6 introduces prior selection and predictive distributions as a means of evaluating priors. Lesson 7 demonstrates Bayesian analysis of Bernoulli data and introduces the computationally convenient concept of conjugate priors. Lesson 8 builds a conjugate model for Poisson data and discusses strategies for selection of prior hyperparameters.

Lesson 6.1 Priors and prior predictive distributions3:46

Lesson 6.2 Prior predictive: binomial example5:20

Lesson 6.3 Posterior predictive distribution4:00

Conoce a los instructores

Herbert Lee
Professor
Applied Mathematics and Statistics

What about after we've observed data?

What's our posterior predictive distribution?

So in this case, we can think about flipping a coin.

We don't know what the probability comes up heads is.

But suppose we observe, after one flip, we got a head the first time.

We want to ask, what's our predicted distribution for

the second flip, given that we saw a head on the first flip?

So in this case, we can think about a posterior predictive distribution,

f of y two, given y one, and so

we can write that as the integral of f of y two, given theta and y one,

times f of theta, given y1, d theta.

Again, we're integrating out theta, but this time,

we're going to use the posterior distribution for theta,

instead of the prior distribution for theta.

In most cases, y2, retreating as independent from y1, so

that will imply that we can simplify this expression, and

just say its y2 given theta, times f of theta, given y1, d theta.

So again, you can see this looks very much like the prior predictive, except we're

using the posterior distribution for theta, instead of the prior distribution.

As a particular example, again, suppose we're thinking of a uniform distribution

for theta, and we observe the first flip, y1 = 1, we get a heads.

What do we predict for the second flip?

This is no longer going to be a uniform distribution like it was in the first

case, because we have some data.

We have one head coming up on the first flip.

So, this gives us some information about the coin.

We're now going to think it's more likely we're going to get a second head, because

it's more likely that theta is at least one half, and possibly larger than a half.

So let's compute this, f of y2, given y1 equals 1,

is equal to the integral, this case, the integral goes from 0 to 1,

of theta to the y2, 1 minus theta, to the 1 minus y2,

times the posterior, which is 2 theta d theta.

We can simplify this a bit, to be, 2 times theta to the y2 plus 1,

times 1 minus theta, to the 1 minus y2, d theta.

We could work this out in a more general form, but

in this case, y2 has to take the value 0 or 1.

The next flip is either going to be heads or

tails, so it's easier to just plop in a particular example.

We can compute the probability, y2 equals 1, given y1 equals 1,

and so here we can just plug in a 1 for y2, and we get the interval from 0 to 1,

of 2 theta squared, d theta, and so, doing that integration,

we get a theta cubed, and this just works out to be two-thirds.

Thus, the compliment, probability of getting a head,

given we saw a tail the first time, is going to be one-third.

So we can see here, that the posterior is a combination

of the information in the prior and the information in the data.

In this case, our prior is like having two data points, one head and one tail.

Saying we have a uniform prior for theta, is actually equivalent

in an information sense, to saying we have observed one head and one tail.

And so then, when we do go ahead and observe one head, it's like we now have

seen two heads and one tail, and so our predictive distribution, our posterior

predictive distribution for the second flip, says, if we have two heads and

one tail, then we have a probability of two-thirds of getting another head, and

a probability of one-third of getting a tail.

Explora nuestro catálogo

Inscríbete de manera gratuita y obtén recomendaciones personalizadas, actualizaciones y ofertas.

Coursera brinda acceso universal a la mejor educación del mundo, al asociarse con las mejores universidades y organizaciones, para ofrecer cursos en línea.

© 2018 Coursera Inc. Todos los derechos reservados.

Descargar en la App Store

Consíguelo en Google Play