Home > Activities > Meetings
2013 French School of Astrostatistics: the regression
The first French school on astrostatistics will be take place near Annecy on Octobre 2125, 2013 :
forum

2013 French School of Astrostatistics: the regression16 April 2013, by vZadLXfmK
CoreyChristian: “the question is not so much whether one believes that the real processes of interest are really computable, but rather whether they are computable in a way that is [] much less complex than their length”Corey: If one has ever used a computer to help one model data, one subscribes to beliefs stronger than this.Christian: “I guess that in most cases this is much less useful than frequentist modelling, which I take as intentionally simplifying things by subsuming all kinds of stuff to ‘random noise’ without the need to worry about computability.”Corey: The prediction error theorem I referred to earlier doesn’t operate in a deterministic environment — it operates in a stochastic environment where the *measure mapping events to probabilities* is a computable function. This includes all practical modelling of any type: if one can write code to simulate it (up to machine precision for an arbitrarily precise machine), then Solomonoff prediction would give (if only one could compute it) a conditional predictive distribution approaching the “true” conditional predictive distribution. And one isn’t limited to deterministic PRNGs in the simulation; one can use genuinely random bits (whatever that means) as input if so desired.

2013 French School of Astrostatistics: the regression13 April 2013, by vlrvLStJNuY
Further to Nonpareil's comment, I think there is a danger in making too much of the "ground game" factor, especially in Quebec. Look at 1984  you had virtual unknown Tory candidates with no organization to speak of overturning Liberal margins of as much as 30,000 votes in some cases. (There's an urban legend that a couple of newlyelected Tory MPs actually showed up for work at the *National Assembly* in Quebec City.) In 1993, the Bloc's very first election, they went from 1 to 54 seats. The provincial ADQ surge in 2007 is yet another example  lots of seats in places where there was little to no ADQ organization. The point is that huge swings in Quebec are not unheard of  organization or no organization. A GOTV makes a huge difference in a dogfight like, say, ParkdaleHigh Park, but in a place where you have the NDP 20 points ahead of the second place party, the lack of GOTV may not matter as much.(Of course, the experience of the ADQ post2007 shows how quickly Quebec voters can tire of the new kid on the block. I suspect that part of the reason the ADQ surge receded almost as quickly as it appeared was that they elected so many 'placeholders' that things quickly turned into amateur hour  the NDP will have to hope it avoids a similar fate.)

2013 French School of Astrostatistics: the regression13 April 2013, by AfkyDWbZjFoM
Further to Nonpareil's comment, I think there is a danger in making too much of the "ground game" factor, especially in Quebec. Look at 1984  you had virtual unknown Tory candidates with no organization to speak of overturning Liberal margins of as much as 30,000 votes in some cases. (There's an urban legend that a couple of newlyelected Tory MPs actually showed up for work at the *National Assembly* in Quebec City.) In 1993, the Bloc's very first election, they went from 1 to 54 seats. The provincial ADQ surge in 2007 is yet another example  lots of seats in places where there was little to no ADQ organization. The point is that huge swings in Quebec are not unheard of  organization or no organization. A GOTV makes a huge difference in a dogfight like, say, ParkdaleHigh Park, but in a place where you have the NDP 20 points ahead of the second place party, the lack of GOTV may not matter as much.(Of course, the experience of the ADQ post2007 shows how quickly Quebec voters can tire of the new kid on the block. I suspect that part of the reason the ADQ surge receded almost as quickly as it appeared was that they elected so many 'placeholders' that things quickly turned into amateur hour  the NDP will have to hope it avoids a similar fate.)

2013 French School of Astrostatistics: the regression13 April 2013, by rjIfzXXdJIzsYl
Further to Nonpareil's comment, I think there is a danger in making too much of the "ground game" factor, especially in Quebec. Look at 1984  you had virtual unknown Tory candidates with no organization to speak of overturning Liberal margins of as much as 30,000 votes in some cases. (There's an urban legend that a couple of newlyelected Tory MPs actually showed up for work at the *National Assembly* in Quebec City.) In 1993, the Bloc's very first election, they went from 1 to 54 seats. The provincial ADQ surge in 2007 is yet another example  lots of seats in places where there was little to no ADQ organization. The point is that huge swings in Quebec are not unheard of  organization or no organization. A GOTV makes a huge difference in a dogfight like, say, ParkdaleHigh Park, but in a place where you have the NDP 20 points ahead of the second place party, the lack of GOTV may not matter as much.(Of course, the experience of the ADQ post2007 shows how quickly Quebec voters can tire of the new kid on the block. I suspect that part of the reason the ADQ surge receded almost as quickly as it appeared was that they elected so many 'placeholders' that things quickly turned into amateur hour  the NDP will have to hope it avoids a similar fate.)

2013 French School of Astrostatistics: the regression8 April 2013, by kzkDCImoQwcT
CONTINUATION OF RESPONSE TO COREY (as these points are general and they keep arising): COREY: For example, in linear regression with a reasonable amount of data, Bayesians can get away with assigning the socalled noninformative prior for the parameters, yielding posterior credible intervals that coincide with frequentist confidence intervals.MAYO: Sure, weâ€™ve noted this before, but this scarcely shows the value/appropriateness of the socalled â€œnoninformativeâ€ prior. As Fisher points out: . (Recall, of course, that the corresponding Bayes test can be wildly different from the credible interval). Thereâ€™s also a difference in interpretation and justification.COREY: I once built a model sufficiently reminiscent of linear regression (actually 3000 linear regressions) that I unthinkingly used the flat prior â€¦. I found that some variance parameters getting stuck close to zero, which messes up the Gibbs sampler. MAYO: Yes, flat priors are scarcely uninformative, in general. Again, no obvious points are scored for this way of proceeding. But I do see your point that youâ€™re able to correct a prior here when you know it makes absolutely no senseâ€¦. Still one might rightly be uncomfortable about using the technique for cases where one was trying to find things out and couldnâ€™t count on â€œbon senseâ€ to rescue one from absurdity.

2013 French School of Astrostatistics: the regression8 April 2013, by jATGimberkHYwlXD
CONTINUATION OF RESPONSE TO COREY (as these points are general and they keep arising): COREY: For example, in linear regression with a reasonable amount of data, Bayesians can get away with assigning the socalled noninformative prior for the parameters, yielding posterior credible intervals that coincide with frequentist confidence intervals.MAYO: Sure, weâ€™ve noted this before, but this scarcely shows the value/appropriateness of the socalled â€œnoninformativeâ€ prior. As Fisher points out: . (Recall, of course, that the corresponding Bayes test can be wildly different from the credible interval). Thereâ€™s also a difference in interpretation and justification.COREY: I once built a model sufficiently reminiscent of linear regression (actually 3000 linear regressions) that I unthinkingly used the flat prior â€¦. I found that some variance parameters getting stuck close to zero, which messes up the Gibbs sampler. MAYO: Yes, flat priors are scarcely uninformative, in general. Again, no obvious points are scored for this way of proceeding. But I do see your point that youâ€™re able to correct a prior here when you know it makes absolutely no senseâ€¦. Still one might rightly be uncomfortable about using the technique for cases where one was trying to find things out and couldnâ€™t count on â€œbon senseâ€ to rescue one from absurdity.

2013 French School of Astrostatistics: the regression8 April 2013
CONTINUATION OF RESPONSE TO COREY (as these points are general and they keep arising): COREY: For example, in linear regression with a reasonable amount of data, Bayesians can get away with assigning the socalled noninformative prior for the parameters, yielding posterior credible intervals that coincide with frequentist confidence intervals.MAYO: Sure, weâ€™ve noted this before, but this scarcely shows the value/appropriateness of the socalled â€œnoninformativeâ€ prior. As Fisher points out: . (Recall, of course, that the corresponding Bayes test can be wildly different from the credible interval). Thereâ€™s also a difference in interpretation and justification.COREY: I once built a model sufficiently reminiscent of linear regression (actually 3000 linear regressions) that I unthinkingly used the flat prior â€¦. I found that some variance parameters getting stuck close to zero, which messes up the Gibbs sampler. MAYO: Yes, flat priors are scarcely uninformative, in general. Again, no obvious points are scored for this way of proceeding. But I do see your point that youâ€™re able to correct a prior here when you know it makes absolutely no senseâ€¦. Still one might rightly be uncomfortable about using the technique for cases where one was trying to find things out and couldnâ€™t count on â€œbon senseâ€ to rescue one from absurdity.

2013 French School of Astrostatistics: the regression5 April 2013, by NJKEKiYuChLpL
One question I have about margins is the degree to which parties in Canada play to win. For instance, John McCain committed considerable resources to Pennsylvania, even though he was well behind, because it was necessary to a victory for him.While Canadian parliamentary elections are not winnertakeall (you can lose but still win more seats), they may be effectively that for some party leaders (Michael Ignatieff is in his 60's and doesn't have a lot of time on his hands). So I guess my point is, what is the appropriate baseline for competitiveness? The ridings that were closest last election? Or the ridings that a party would need to win in order to win government. So yes, you are probably going to send GOTV resources to the marginals as the endgame approaches, but before then in picking a strategy, you have to have to be thinking about ridings like Guelph  even if you are Harper and its 2004 (the Tories lost there by 19 points). Testable hypothesis: do young and secure leaders target marginals while older and more vulnerable (to losing party leadership) leaders target what I will call linchpin ridings (the ridings the party would need to win in order to form a government).

2013 French School of Astrostatistics: the regression28 March 2013, by hDIcQMYWcxFXrsToq
First good health shouldn't be a fad nor done alone. Gather your family & talk about how health matters & realize change will be challanging but together anything is possible. Together come up with a plan for moving more & eating less junk. Think about going online or to the library to do some recipe reading. Being spring its a good time to think about starting a garden, even a sunny window means fresh herbs. Invest in good cooking tools make it fun by exploring the new lines. Get measuring spoons, cups & a scale to learn what a real serving is (often less than you may think). Take a field trip to the market & discover your inner detective. Start by trying lighter favorite foods, key words less fat, baked, reduced, lower sugar/salt. Tip manufactures can vary greatly read the label before choosing between products. Try to stay natural if you can't pronounce it do you really want to eat it? Try to drink more water (keep a cold pitcher with fruits or sliced cucumber always on hand) save real fruit juices & milk for meals (with kids this can help with the budget). As with anything be flexible & enjoy life, you don't have to give up foods you love moderation & don't mindlessly reach instead really savor the moment.

2013 French School of Astrostatistics: the regression25 March 2013, by TOlQOqQpSuHxxmdZbzg
Corey“What is being checked: whether you adequately captured your prior degrees of belief in a claim, hypothesis or model?”Moreorless. I would phrase it as checking whether I adequately encoded my available prior information. I hold, contra Kadane, that it’s not about my belief — it’s about the degree of plausibility justified by a given state of information. In principle, any two agents with exactly the same prior information should end up with the same prior distribution.“If the prior is changed by data, how can it represent your beliefs based only on info prior to the data? Or do you not get to change your prior if you donâ€™t like the posterior?”Since assigning priors by introspection is errorprone, one useful insight is that it makes little material difference whether a given piece of information is in both the prior and the likelihood or just in the likelihood. For example, in linear regression with a reasonable amount of data, Bayesians can get away with assigning the socalled noninformative prior for the parameters, yielding posterior credible intervals that coincide with frequentist confidence intervals. In that analysis the prior on the logarithm of the variance of the noise is uniform over the whole real line, which corresponds to no reasonable prior state of information, e.g., we know in advance that our timing devices don’t have measurement noise on the order of the Planck time unit nor on the order of hours. This prior works anyway because the data themselves rule out ridiculously precise or imprecise measurement noise variance; encoding that information in the prior would make no material difference to the subsequent inferences.I once built a model sufficiently reminiscent of linear regression (actually 3000 linear regressions) that I unthinkingly used the flat prior on the logvariance parameters. Upon running my Gibbs sampler to generate samples from the posterior distribution, I found that some variance parameters getting stuck close to zero, which messes up the Gibbs sampler. There’s a technical fix for this, Gelman’s parameterexpanded Gibbs sampler, but I realized the source of the problem was that neither the improper prior density I had used nor the likelihood ruled out effectively infinitely precise measurements for some of the “regressions”. But I had prior information about the expected magnitude of the measurement error from the literature and also directly from the analytical chemist who collected the data. So I switched to a different prior density, one that encoded information ruling out ridiculously high precision. This fixed the glitchy Gibbs sampler and also materially improved the resulting inferences.So it was not the case that the prior information was changed by data, nor was it the case that I couldn’t change the prior density when I found features of the posterior inconsistent with the prior information.

2013 French School of Astrostatistics: the regression20 March 2013, by VTCOgDYtLJfZUqDepHb
RMhFCY doymurjpsovs

2013 French School of Astrostatistics: the regression20 March 2013, by coXCYKiVUpKPElNdo
Luke,Funny you mention that, when I first stetrad looking into it I noticed a lot of bad press regarding the name. At the time, I didn’t look any further and thought the liklihood that it’d be the same person were slim. After reading your thoughts on the matter though, I think you could have very well hit the nail on the head. You have to admire the tactics used to remove the bad press though, heh.Al.