Alex,
I really like the presentation, perhaps that's because I agree fully on the approach and it supports the claims that I have made on the impossibility of having anything genuinely better than the Bayesian approach. The presentation tells very clearly, why the Bayesian approach is difficult and why it doesn't provide objective results, but it explains also, why no genuinely better alternatives exist.
I found the second half of the presentation more interesting starting from slide 67. That starts with
<blockquote>
Basic principle: it is always better to recognise than to ignore uncertainty, even if modelling and analysis of uncertainty is difficult and partial.
</blockquote>
That's a staring point that I have also used to defend the Bayesian approach. Many other approaches start by formally accepting ignorance, but proceeding then in a way, where the actual significance of both the ignorance and of what we know anyway, are not faced explicitly, but taken care by the methodology in a too rough way. The problem that I have we some of the approaches proposed by Judith is that they lose information that way have anyway and therefore make the uncertainty to be a bigger monster than it really is. Only a serious attempt to apply the Bayesian approach can tell more about the limits of uncertainty.
Understanding the nature of the three stages listed on slide 98 is also very important in that approach. The main problem may then be in valuing the Stage 2. The whole IPCC approach is closely related to Stage 2. Understanding, what's the value of, what the presentation calls <i>a scientific Bayes analysis</i> becomes important. The name <i>scientific Bayes analysis</i> is a bit questionable, but in a sense all scientific knowledge can be considered as outcome of such <i>scientific Bayes
analysis<i>, although it may get gradually closer to the third stage of <i>an objective Bayes analysis<i>.
↧