[ppml] Combining Forecasts

[Marshall Eubanks]

On Aug 30, 2007, at 12:31 PM, David Conrad wrote:

>> If we are going to use the forecasts to incent people to move from
>> IPv4
>> to IPv6 then we should be as accurate as possible.
>
> It is a bit challenging to model human behavior.
>
> How do you model a "run on the bank"?

Modeling it is possible; predicting it is almost impossible until it  
starts.

What you could maybe do is predict the probability of it happening.

It would help to have data to compare against, but I cannot really  
think of any suitable data or a previous case of a run on a registrar.

There is of course a huge literature on financial panics - they seem  
to be currently modeled by Markov switching models,

see
http://www.fdic.gov/bank/analytical/working/wp2006_01/wp2006_01.pdf
http://www.cambridge.org/resources/0521790182/1554_117089.ppt
http://welch.econ.brown.edu/cascades/

How applicable this is is of course unclear.

I must say that with the resources available I am not too sanguine  
that we would get useful predictions even of probabilities.

Regards
Marshall


>
> How do you model a transition of public space to NAT'd space as an
> effort towards conservation?
>
> Projections such as Geoff's and Tony's are useful, but I do not
> believe they are something you should bet your house on.
>
> Regards,
> -drc
>
> _______________________________________________
> PPML
> You are receiving this message because you are subscribed to the  
> ARIN Public Policy
> Mailing List (PPML at arin.net).
> Unsubscribe or manage your mailing list subscription at:
> http://lists.arin.net/mailman/listinfo/ppml Please contact the ARIN  
> Member Services
> Help Desk at info at arin.net if you experience any issues.