tag:blogger.com,1999:blog-8666091.post114523037207628351..comments2021-06-11T20:12:08.082+02:00Comments on The Reference Frame: Stanislav Petrov supersedes Easter Bunny and Jesus ChristLuboš Motlhttp://www.blogger.com/profile/17487263983247488359noreply@blogger.comBlogger3125tag:blogger.com,1999:blog-8666091.post-1145354335009406482006-04-18T11:58:00.000+02:002006-04-18T11:58:00.000+02:00Dear Lumo,I agree. Not only the prior is a gamma -...Dear Lumo,<BR/><BR/>I agree. Not only the prior is a gamma -distributed, they also model one out of two "evidence distributions" with a gamma distribution...<BR/><BR/>Although I'm not a big fan of Bayesian analysis myself, I do find James's contribution (the idea of using Bayes's rule for determing climate sensitivity), per se, interesting. However, what does disturb me is the actual contents of the paper, and the fact, it has been accepted to, what I have understood, a major journal in the field. If I had been a reviewer, I would have demanded a major rewrite, see below.<BR/><BR/>Not only the choice of gamma distribution is somewhat wierd, the main problem, in my opinion, in the paper is that there is no real discussion of the fitting of the "evidence distributions". Usually the likelihood is obtained from data, but now since it was determined otherwise, there should have been some evidence that the fitting is adequate. I gave James a link to families of distributions, which are rather common for modelling skewed distributions (as he obviously wants to have some skewed distributions). If you had several "evidence distributions", then the fitting is not so important, but they only had two!<BR/><BR/>The other thing I do not understand in the paper, is the starting point. Why do they have "20th century warming" as a prior, shouldn't it be more like an evidence distribution? There exists a rather obvious (intuitively appealing) prior:<BR/>calculate the temperature change with an assumption that everything except CO2 is kept fixed. I'm no physicist, but I've understood that this is rather straightforward (hint: it would be nice to see this calculation, e.g., in your blog ;). Then your prior is simply a Gaussian with this mean, and some variance of your choice. This, at least in my opinion, reflects the idea of an uninformative prior for climate sensitivity: since we do not know, a priori, what is the net effect of the forcings, it is best to model the situation with Gaussian such that the mean is the situation with no feedback effects at all.<BR/><BR/>Well, "climate science" is not my field, and James (who also seems to be reading this blog ;) did give me an answer (which was pretty much I expected him to say), so I will not bother with this issue any more.<BR/><BR/>Cheers, pikkupoikaPikkupoikahttps://www.blogger.com/profile/15860421173496735082noreply@blogger.comtag:blogger.com,1999:blog-8666091.post-1145297671201780062006-04-17T20:14:00.000+02:002006-04-17T20:14:00.000+02:00Dear Pikkupoika/Jani,I certainly agree with you th...Dear Pikkupoika/Jani,<BR/><BR/>I certainly agree with you that assuming that there is a strict 0% probability that the climate sensitivity is negative is a manifestation of a bias of the scientist, and I am not aware of any solid argument why it must be positive.<BR/><BR/>More generally, it seems bizarre to assume a gamma distribution for a climate sensitivity. A gamma distribution is a derived concept generalizing events distributed according to the Poisson distribution (and the variable we are distributing is essentially time) - what does it have to do with climate sensitivity? Kind of weird. ;-)<BR/><BR/>Best<BR/>LubosLuboš Motlhttps://www.blogger.com/profile/17487263983247488359noreply@blogger.comtag:blogger.com,1999:blog-8666091.post-1145296048402148272006-04-17T19:47:00.000+02:002006-04-17T19:47:00.000+02:00Sorry, this has nothing to do with the post. I jus...Sorry, this has nothing to do with the post. <BR/><BR/>I just wanted to let your know that I found something, at least in my opinion, "curious" in J. Annan's recent paper. I (with pseudoname "Jani") submitted the question regarding that to the end of his <A HREF="http://julesandjames.blogspot.com/2006/04/subjectivity-objectivity-and-belief-in.html" REL="nofollow">recent post</A>. Since I do not want to be identified, and I spent some time on the issue, I just want someone "with undestanding" to to know about the "curiosity" (if he decides to delete the comment or something). So please check it.Pikkupoikahttps://www.blogger.com/profile/15860421173496735082noreply@blogger.com