Rmetrics - timeDate Class

Sun, Jun 13, 2004 11:37 AM
I would like to clarify my suggestion.

My suggestion was simply how to implement the current functionality
without making the user set environment variables.

Whether time zones are important or not does not seem to be 
relevant to that discussion since I was not proposing to change
any outward functionality -- just how to implement the existing
functionality in more convenient way.



Date:   Sun, 13 Jun 2004 15:03:22 +0000 
From:   Diethelm Wuertz <wuertz@itp.phys.ethz.ch>
To:   <ggrothendieck@myway.com>, <r-sig-finance@stat.math.ethz.ch> 
Subject:   Rmetrics - timeDate Class 

 
Many thanks for your suggestion,

I don't know if your suggestion works. I had many problems in 
calculating time
properly and therefore I decided to do it the way I did it, in GMT + 
FinCenter
information. Here comes an example:

In the eighties a trader in Frankfurt is awating the closing prices from 
the Zurich
exchange. When gets he or she the information in local time? It's not 
the same time
in Frankfurt as in Zurich although both cities belong to the same time 
zone! Since
in the eighties Zurich and Frankfurt moved to DST at different dates 
there is a
difference of one hour for some dates!

I know of some other critical examples, and so I decided to work 
internally with
GMT excluding all time zone information. The problem is that Zurich and 
Frankfurt
cannot be distinguished by the time zone alone. One needs more 
information! So
I use for both Zurich and Frankfurt GMT with the (different) DST info 
available from
the timezone database. Otherwise I have to introduce DST-MET (Middle 
European
Time) in several ways for different cities in Europe.

Maybe there is a better solution to such problems ...

The whole timedDate class still needs much more testing ... (I did it to 
some extent
under MS Windows)

I put this on the R-sig-finance discussion list, because I think these 
questions
are of more general interest.

Regards

Diethelm
... wrote: