date: Wed, 21 Aug 2002 13:27:04 +0100
from: Benito Mueller <benito.muelleratXYZxyzlosophy.oxford.ac.uk>
subject: Re: Impact figure projections
to: Mike Hulme <m.hulmeatXYZxyz.ac.uk>
many thanks for taking the trouble. I will definitely take you comments
into consideration for future presentations of the result (e.g. at
COP8). Please find attached a substantially revised version (including
some further sensitivity analyses, on Tom's suggestion). I also included
some more ENSO 'stuff', which you may be interested in (a physicist friend
helped me out with some Fourier analyses)
As for your point (3), I'm afraid I do not have the capacity for this but
would be willing to help someone who does.
Again, many thanks
At 12:58 21/08/2002 +0100, you wrote:
>These comments are probably too late for you but holidays have intervened.
>1. You should really explore how sensitive your model is to different
>starting dates. If you chose 1980 instead of 1975 then all your
>predictions would be less extreme and if you started in 1987, your model
>would predict reductions! These sort of regression based models can be
>very sensitive to the particular year you chose to start from.
>2. You should also explore how sensitive the model predictions are to your
>preferred use of % units (i.e., of total population) rather than absolute
>numbers of people. One could argue that absolute numbers is a more
>sensible unit to use - although I admit there are arguments either
>way. How sensitive is your model to this?
>3. Why do you not set up a direct step-wise multiple-regression model
>with three predictors - time, global T and an ENSO index? Thus you tackle
>all three issues simultaneously. Different future ENSO indices can be
>used to examine sensitivity then.
>4. And of course the general point is to emphasis the need to address the
>non-climatic part of the trend just as much - if not more - than the
>climatic part of the trend (the latter simply provides the entry point
>into the UNFCCC world).
>Other comments - not sure what you mean in your use of the word
>'significant' - this can sometimes have a formal statistical meaning but
>it also has a popular meaning. Be careful how you use this.
>Also, once or twice you insert value statements e.g. a 'truly worrying
>trend' which I would have thought are best left out.
>I'd be interested to see how this discussion develops.
>At 14:09 30/07/02 +0100, you wrote:
>>as promised (threatened?), please find attached a short chapter on the
>>'Demand-side' for Impact response measures. It's primary aim is to
>>highlight the urgent need to consider vulnerability reduction and impact
>>response measures. I take it from our conversation here in Oxford some
>>days ago that you concur with the need to highlight these issues. Given
>>this, I was hoping whether you and/or the colleague you mentioned could
>>have a quick look at the chapter and tell me where I am wrong and how I
>>could improve the argument.
>>As I will be giving a presentation of this and other material to Swiss
>>government officials and representatives from the International Red Cross
>>and the UN Office for the coordination of Humanitarian Affairs in two
>>weeks time, I would be most grateful if you could indicate your key
>>Dr Benito M�ller
>>+44 (0)1865 889 135 (direct)
>>+44 (0)1865 311 377
Attachment Converted: "c:\eudora\attach\Mueller II.a V2.pdf"
Dr Benito M�ller
+44 (0)1865 889 135 (direct)
+44 (0)1865 311 377