Bayesian linear regression

Bayesian linear regression using the standard noninformative prior. Although the marginal posteriors for the regression coefficients and the variance are available in closed form, we simulate from the posterior distribution using a QR decomposition of the design matrix for numerical stability and efficiency.

[test-mmbv]

i've heard that one of the greatest advantages of bayesian linear regression is more responsive updates to new data.
how is new data incorporated into the regression formulation to update estimated betas?

[FlareSoul]

I'm definitely in the wrong class. This guy must be teaching to Ph.D's students

[grupobits]

Sorry teacher this is too much jargon for me i hope to understand this easily someday with a lot of effort and discipline

[toastersman]

Do you have a R example with the presence of a correlation matrix for the variance of the errors?

[ccuuttww]

Hey thanks for this tutorial but I want to discuss the inverse gamma function
If u use 1/rgamma(n, alpha, loc=0, scale=beta) in R
u will get slightly different between python

In python I use
alpha = df
beta = df * s2
sigma = scipy.stats.invgamma.rvs(alpha, loc=0, scale=beta, size=10000)
and the result as follow

Output:

R:
mean(sigma)
1.047957
mean(beta[, 1])
-0.5788454
mean(beta[, 2])
0.2152288

Python:
np.mean(sigma)
1.0511557478134843
np.mean(beta[:, 0])
-0.6404757705536689
np.mean(beta[:, 1])
0.21606140101561436

[AP-vuid]

Since X and y are given, we can compute vector betta-hat, matrix V-betta and scalar s^2. Then sigma square and betta become defined by distributions and are random variables. These two variables are plugged into multinomial natural distribution for y and this is how we obtain our wanted answer. I have two questions: 1. Is my above statement correct? and 2. Are you really a professor, the explanation sounds like a lecture of a cab driver, you messed everything, did not articulate where is vector, scalar or matrix what is computed value and what is random variable.

[lovemormus]

This is just a ordinary course given in the schools, where the speaker doesn't actually explain the essence of Bayesian regression, nor the applications of BLG. Thus, students would never be interested in this topic. Should vote down this video.