Robin's Blog - Latest Commentshttp://robinceblog.disqus.com/enTue, 07 Jan 2014 18:03:12 -0000Re: Blog of Robin Incehttp://www.robinince.net/blog/2013/03/21/saving-large-files-efficiently-in-matlab#comment-1191337056<p>Hi Robin, <br>the savefast.m function is awesome. Thanks. <br>Is there a way of make it case sensitive? in line 66 the h5create function is not case sensitive is there a way to change that?</p><p>thanks,<br>Jose Herrero<br>CU @ NYC</p>Jose HerreroTue, 07 Jan 2014 18:03:12 -0000Re: Blog of Robin Incehttp://www.robinince.net/blog/2013/04/18/fitting-multivariate-t-distributions-to-data#comment-1156273442<p>My first comment, thanks!</p><p>I think the standard way in general is maximum likelihood. My package here implements maximum likelihood for the t-distribution (which unlike for Gaussian doesn't have a closed form solution which can be directly evaluated). I think the algorithm I implemented here is as close as anything to a standard for t-distributions - I just couldn't find an existing Matlab implementation. MAP is a way to get a point estimate within a Bayesian framework - so requires priors over all parameters. And if you have a full Bayesian posterior it is probably better to keep it around and do further statistical analysis over the full distribution of possible parameters rather than a point estimate.</p>robinceMon, 09 Dec 2013 14:07:33 -0000Re: Blog of Robin Incehttp://www.robinince.net/blog/2013/04/18/fitting-multivariate-t-distributions-to-data#comment-1156259409<p>Cheers. What is the standard way that people fit such distributions to data? I suppose maximum a posteriori estimation (?). Is your method an M.A.P.?</p>SohailMon, 09 Dec 2013 13:56:19 -0000Re: http://www.robinince.net/blog/2011/12/19/first!http://www.robinince.net/blog/2011/12/19/first!#comment-389957119<p>And a comment.</p>robinceMon, 19 Dec 2011 12:06:11 -0000Re: http://www.robinince.net/blog/2009/08/29/syntax-highlight-testhttp://www.robinince.net/blog/2009/08/29/syntax-highlight-test#comment-388187546<p>Posting anon</p>AnonFri, 16 Dec 2011 14:34:17 -0000