![I came across this question about truncated Normal simulation on X validated, pointing out at a 1963 [AX II!] Boeing technical report by George Marsaglia! The proposal uses a translated Exponential variate with scale 2, and then takes the squared root. This is an alternative of my (1995) [and John Geweke's (1993)] proposal using a translated Exponential variate with optimised scale and I wonder which solution is better. Presumably Marsaglia's, especially since the scale of 1/2 is the optimal one! Fodder for the next Monte Carlo exam, possibly, if my students do not come across this 'Og entry.](https://i0.wp.com/xianblog.fr/wp-content/uploads/2024/09/temp-4.png?resize=450%2C340&ssl=1)
Related
This entry was posted on September 23, 2024 at 12:24 am and is filed under Books, pictures, Statistics, University life with tags accept-reject algorithm, Boeing, cross validated, George Marsaglia, John Geweke, Stack Exchange, truncated normal. You can follow any responses to this entry through the RSS 2.0 feed.
You can leave a response, or trackback from your own site.
October 2, 2024 at 1:52 am
Interesting — on a quick look, this seems to be a rejection sampler?
On a related note, we recently came up with a neat trick for doing blocked Gibbs sampling that involves high-dimensional truncated Gaussian. (pg. 12-13 https://arxiv.org/pdf/2309.09371 )
October 2, 2024 at 11:40 am
Yes, this is another accept-reject sampler. Thanks for the pointer to your paper.