We are proposing the following changes to the calculation of NG correlations:

1. weight the data set (3 calendar months) used in calculating correlations 
(most recent data weighed heavier)
2. use respective contract prices, instead of prompt month prices (i.e. for 
Nov-00 correlations use Nov contract prices for the last 3 months, as opposed 
to prompt month prices for the last three months.


Tanya,

I have confirmed with Ted and he gave us green light to make both changes. 
Did we get an opinion from Vince?

Winston,

It is my understanding, that this changes apply to NG correlations only, not 
the correlations between commodities. We will test the changes in gas and 
then decide on world-wide implementation. Any estimate on timing of this 
implementation?

Cassandra,

Ted suggested that you and Veronica should document this as a change in VaR 
parameters and inform all commercial desk heads of these changes. We intend 
to make them for NA gas first, but ultimately make these changes consistent 
across all commodity groups. Let me know if you have questions.

Thanks, Vlady.





Wenyao Jia
10/13/2000 03:43 PM
To: Vladimir Gorny/HOU/ECT@ECT
cc: Tanya Tamarchenko/HOU/ECT@ECT, Jin Yu/HOU/ECT@ECT 
Subject: Re: VAR calibration issues  


Vlady,

Also in the meeting, we identified that there are still some issures 
regarding to the correlation matrix calculations.

Since different commodity has different expiration dates.  When calculate 
correlation between two commodities, the two may have different prompt 
months.  Are we going to use prices on two different prompt months OR are we 
going to use the prices on the SAME month disregarding prompt months.

Because above issues, Jin is not going do any changes on the correlation 
matrix calculation until above issures can be solved.

Thanks!

Winston






Tanya Tamarchenko
10/13/2000 03:16 PM
To: Vladimir Gorny/HOU/ECT@ECT
cc: Wenyao Jia/HOU/ECT@ECT, Jin Yu/HOU/ECT@ECT, Jin Yu/HOU/ECT@ECT 
Subject: Re: VAR calibration issues  

Vlady, we met with Winston and Jin today regarding VAR calibration issues.

The outcome on this discussion is:
1. Jin will put weights into calculation of factor loadings;
2. Jin will change the way factor loading are calculated. For each commodity 
the prompt month contract will be selected for the effective date of vatrfacs 
run.
Then the historical prices will be collected for 3 month for all forward 
contracts starting from
selected prompt month contract. The variance-covariance matrix will be 
calculated
based on these data, it will be converted into correlation matrix, then 
factor loadings 
analysis will be performed on the correlation matrix.


Tanya.