1,725 reputation
513
bio website
location Ottawa, Canada
age
visits member for 2 years, 6 months
seen Sep 12 at 1:03

Corey Yanofsky, Ph.D.

I'm a biostatistician with an interest in financial statistics located in Ottawa, Canada. Philosophically, I'm a Bayesian of the Cox-Jaynes school. In practice I'm a statistical ecumenist whose role models are Michael I. Jordan and Andrew Gelman.


Apr
25
awarded  Enlightened
Apr
25
awarded  Nice Answer
Mar
9
awarded  Yearling
Feb
22
comment On the tractability of posterior distributions
Chamberlain Foncha, if you have the covariance matrix, you just calculate a matrix square root (using the Cholesky decomposition, or if $N$ is too big for that to yield stable results, the eigendecomposition) and then use a matrix-affine transformation of a product of standard normal random variates. This will work until $N$ is so big that the eigendecomposition computation outstrips one's computing resources.
Jan
30
comment Uniform Random on $(-\infty,\infty)$
Here's a relevant draft article from the arXiv.
Jan
27
asked Monotone likelihood ratio property: check my proof; also, who proved it first?
Jan
25
awarded  Nice Answer
Jan
11
awarded  Nice Question
Sep
30
awarded  Scholar
Sep
30
accepted What is the decision-theoretic justification for Bayesian credible interval procedures?
Sep
28
comment How do programs like BUGS/JAGS automatically determine conditional distributions for Gibbs sampling?
@Glen If you provide an example that's given you difficulty, I'll do the inspection.
Sep
26
comment Bayesian analysis problem
Your data are non-negative ratios, so take the logarithm and model on $\mathbb{R}$. A non-parametric method will probably serve you well. You might need to mix in a point mass at 0% (on the original scale, corresponding to $-\infty$ on the log scale).
Sep
25
comment What is the difference between posterior and posterior predictive distribution?
That posterior predictive distribution graph needs new axis labels and a caption or something. I get the idea because I know what a posterior predictive distribution is, but someone who's just figuring it out could get seriously confused.
Sep
24
comment Why does Bayes' Theorem work graphically?
Try this.
Sep
23
comment Running regularized logistic regressions on very large datasets
Hmm... biglars uses the transparent-flat-file-storage package ff to get around that difficulty. Maybe we need a bigglmnet package. I don't know your R hacking skill level; it's within my capabilities, but I have other things on my to-do list right now...
Sep
20
comment What's the bayesian equivalent of a general goodness of fit test?
Something like this might fit the bill.
Sep
20
comment Running regularized logistic regressions on very large datasets
glmnet
Sep
13
comment Convenient posterior distribution for homogeneous bivariate Gaussian model
@StéphaneLaurent I'm pretty sure it does require optimization. My experience is limited to reading the documentation for and papers describing the hoa S-PLUS library; the optimization code is built-in.
Sep
12
comment Convenient posterior distribution for homogeneous bivariate Gaussian model
@StéphaneLaurent I propose profile likelihood approaches because they target frequentist coverage directly (and I presume you either have no relevant prior information or you want to exclude any such information from the analysis). Modified profile likelihood applies a third-order correction to the signed profile-log-likelihood-ratio test statistic, making its sampling distribution closer to normal. Google has lots of good stuff. (I hope your correlation coefficient is positive...)
Sep
12
comment Convenient posterior distribution for homogeneous bivariate Gaussian model
@StéphaneLaurent Is there a one-to-one transformation between, say, $(\mu_1,\mu_2,\theta)$ and $(\mu_1,\mu_2,\sigma)$? If so, then I still say modified profile likelihood. It'll take brain power, be computationally intensive, and you'll want to check coverage by simulation, but it will work.