Statistics slides (3)
Here is the third set of slides for my third year statistics course. Nothing out of the ordinary, but the opportunity to link statistics and simulation for students not yet exposed to Monte Carlo methods. (No ABC yet, but who knows?, I may use ABC as an entry to Bayesian statistics, following Don Rubin’s example! Surprising typo on the Project Euclid page for this 1984 paper, by the way…) On Monday, I had the pleasant surprise to see Shravan Vasishth in the audience, as he is visiting Université Denis Diderot (Paris 7) this month.
Xi'an's Og 
October 20, 2014 at 4:52 pm
I’m starting to think that Rubins (1984) way of describing Bayesian statistics is the most intuitive explanation. If I ever wrote a tutorial on Bayes I would surely start with his explanation.
October 20, 2014 at 8:13 pm
I still have to write the slides for the Bayesian section so I may follow your advice. Good job with the socks, by the way. Could you extend to the problem of finding the missing ones?!
October 20, 2014 at 9:11 pm
The socks were fun! ABC is one of those techniques that I really wished that somebody told me about earlier. Just like I wished somebody had told me about bootstrap before I had to go through the t-test/chi-square/anova dance that is psychology statistics. Ah, the missing ones! Would be difficult to estimate without extensive data regarding a persons geographical location over time. Would it be easier if the socks were missing at random? :)
October 20, 2014 at 11:41 pm
Can I ask, since I use an “exact rejection sampling” algorithm in the socks blog post (that is, I sort of use the “identity function” as the summary statistics) is it still ok to call it approximate bayesian computation? (which I do call it…) It is still likelihood free, at least.
October 21, 2014 at 1:50 pm
Yes, this is “exact ABC” since you wait for the data and pseudo-data to be equal.
October 21, 2014 at 2:22 pm
Exact Approximate Bayesian computation :) Thanks!
October 11, 2014 at 2:41 pm
I should mention that Christian’s third year undergraduate lecture, which I attended last Monday, was about at the right level (just beyond my level) for me. I guess the MSc level courses at Sheffield are really focused more on delivering practical skills. Or maybe the real hard work happens at the undergrad level in Statistics? I think that is true for computer science. When I was a grad student at Ohio State, I had to do a lot of really difficult courses at the undergrad level in computer science before I was allowed to do the MS level courses, which were at a much higher level (hardly any programming in comparison to the undergrad classes).
I’m looking forward to the next lectures in this course.