Re: [amibroker] New Question on Watchlist Averages

[dan_public@xxxxxxxxxxx]

I agree 100% with TJ!   VERY strongly recommended.  I have five scans that create and/or use ATC composites.   Total run time is about 45 minutes per day.   However, when I display the ATC symbols or use them in further analysis, response time is very fast.  

Note that this is a standard Data Warehouse methodology.  Pre-aggregation and pre-processing of data can take a long time, but the user sees the processed results very quickly.

Regards,

Dan.

-------------- Original message -------------- 
From: "Tomasz Janeczko" <groups@xxxxxxxxxxxxx> 
Hello,

You would need to check for Null values and/or use Nz (null to zero) because of the fact
that if you don't have data for particular security Foreign() will give you null values at the beginning of the array.

Personally I would advice using AddToComposite in a regular way (scan over entire database).
This way you would arrive to correct results faster and with less code. This is so because plain scan
with addtocomposite does not involve any loops therefore it is fast even if larger number of symbols is
being analysed. Especially once your needs grow over time and you will want to calculate averages
not only for few watchlists but also for some industry groups, etc.

Best regards,
Tomasz Janeczko
amibroker.com
----- Original Message ----- 
From: Ken Close 
To: amibroker@xxxxxxxxxxxxxxx 
Sent: Wednesday, August 09, 2006 8:44 PM
Subject: [amibroker] New Question on Watchlist Averages


I have been making some progress using this code, I think from the help files and/or mentioned here on the list:

function CreateAverageForWatchList( listnum )
{
   // retrive comma-separated list of symbols in watch list
   list = CategoryGetSymbols( categoryWatchlist, listnum );

   Average = 0; // just in case there are no watch list members

   for( i = 0; ( sym = StrExtract( list, i ) ) != ""; i++ )
   {
      f = Foreign( sym, "C" );
      if( i == 0 ) Average = f;
      else Average = Average + f;
   }
   return Average / i; // divide by number of components
}

I am creating averages of various watchlists and it is working....except for.....

The resulting average is no longer than the watchlist member with the least or shortest amount of data.
If I have 20 members, 19 of which have data going back into the early 90s but 1 member has data for the last 6 months, then the average which is created is only 6 months long.

Is there any modification to the above code such that an average for each date will be created from which ever members have data for that date.  In my example, I would get an average based on 19 members up until 6 months ago whereupon the average would be based on 20 members.

Can this code do that? If not, how could the code be altered to accomplish that.

Thanks to all who are trying to help me with this. Some of the other suggested approaches, using jscript (thanks Dan), etc, have been a little daunting and I was able to get the above to work until I discovered the truncated data (shortened average to the shortest list member).

Ken