Academic humor
Apr. 18th, 2012 05:05 pm![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png){kind=small}
eve11On the discussion of the paper "Efficient Construction of Reversible Jump Markov Chain Monte Carlo Proposal Distributions" by S. Brooks, P. Giudici, and G. Roberts. Journal of the Royal Statistical Society, Series B (Statistical Methodology), Volume 65, No. 1 (2003).
Second discussant: Xiao-Li Meng (Harvard, Cambridge and University of Chicago) (and yes, of Meng and Rubin fame):
Initially I was a little puzzled at being invited to discuss this paper, as I could not locate my name anywhere in it. Then I realized that this perhaps was intentional, for the second discussant is often perceived to be the 'bad guy'--an 'uncited' discussant tends to be more critical. As a 'good guy', one often gets away with saying, 'The authors are to be congratulated for a very interesting paper. Now look at what I have done.' Being a bad guy, however, one actually has to read the paper. I read every word (at least through the first page), partially in the hope of finding flaws to be labelled 'rubbish', a word, as I understood which was previously used in such a context. Unfortunately for me as the second discussant, but fortunately for readers, the only place where I could possibly do so is at the third sentence of the paper. This sentence, if read literally, suggests that there is a Bayesian model determination problem beyond the choice of prior and the specification of the likelihood. What could that be? Of course it is unfair to blame the authors for this common misuse of 'model determination', which should be 'submodel determination'. [etc. on some interesting if semantic points].
Having done my duty as the bad guy, let me return to my normal role as a good guy so that I can justifiably advertise some of my own work. [...snip 2 long paragraphs on the similarities between auxiliary variables and data augmentation, and a method of matching spread through something called warping, of which there are versions I, II, and III]. Of course, this is speculative, but my positive experiences with warp III gives me some hope that transdimensional warping is not science fiction.
To echo my initial 'complaint' about the lack of citation, let me end my discussion with a citation ranking in Science Watch (May-June 2002). Among the top 25 ranks of mathematicians worldwide in terms of papers published and cited during 1991-2000, statisticians occupied 19 spots! Parts of the list reads like a 'who's who?' for the Markov Chain Monte Carlo club, including two of the Society's Presidents, Adrian Smith (number 3) and Peter Green (number 13). I certainly hope that this paper will help to increase Gareth Roberts's current ranking (16), and, with a little help from the authors, my own ranking would be moved up as well!
With that thought, it gives me great pleasure to second the vote of thanks for this potentially highly cited paper!
Second discussant: Xiao-Li Meng (Harvard, Cambridge and University of Chicago) (and yes, of Meng and Rubin fame):
Initially I was a little puzzled at being invited to discuss this paper, as I could not locate my name anywhere in it. Then I realized that this perhaps was intentional, for the second discussant is often perceived to be the 'bad guy'--an 'uncited' discussant tends to be more critical. As a 'good guy', one often gets away with saying, 'The authors are to be congratulated for a very interesting paper. Now look at what I have done.' Being a bad guy, however, one actually has to read the paper. I read every word (at least through the first page), partially in the hope of finding flaws to be labelled 'rubbish', a word, as I understood which was previously used in such a context. Unfortunately for me as the second discussant, but fortunately for readers, the only place where I could possibly do so is at the third sentence of the paper. This sentence, if read literally, suggests that there is a Bayesian model determination problem beyond the choice of prior and the specification of the likelihood. What could that be? Of course it is unfair to blame the authors for this common misuse of 'model determination', which should be 'submodel determination'. [etc. on some interesting if semantic points].
Having done my duty as the bad guy, let me return to my normal role as a good guy so that I can justifiably advertise some of my own work. [...snip 2 long paragraphs on the similarities between auxiliary variables and data augmentation, and a method of matching spread through something called warping, of which there are versions I, II, and III]. Of course, this is speculative, but my positive experiences with warp III gives me some hope that transdimensional warping is not science fiction.
To echo my initial 'complaint' about the lack of citation, let me end my discussion with a citation ranking in Science Watch (May-June 2002). Among the top 25 ranks of mathematicians worldwide in terms of papers published and cited during 1991-2000, statisticians occupied 19 spots! Parts of the list reads like a 'who's who?' for the Markov Chain Monte Carlo club, including two of the Society's Presidents, Adrian Smith (number 3) and Peter Green (number 13). I certainly hope that this paper will help to increase Gareth Roberts's current ranking (16), and, with a little help from the authors, my own ranking would be moved up as well!
With that thought, it gives me great pleasure to second the vote of thanks for this potentially highly cited paper!
