|
PureBytes Links
Trading Reference Links
|
I respectfully disagree. If you are not concerned about bar-by-bar accuracy
than you are correct. In that case the "1" method works fine. As long as you
know that this method will pick up holes of several days but that it will
not pick up single bar holes.
If you want to perform bar-by-bar analysis then it doesn't work. Please have
a look at the attached capture that shows how the "1" method fails and
misses a lot of holes.
When you are working with a low stock count than this can give serious
errors. The omission of holes shown in the attached capture would result in
several counting errors of about 20%.
Best regards,
Herman.
> -----Original Message-----
> From: dtsokakis [mailto:TSOKAKIS@x...]
> Sent: Sunday, April 28, 2002 6:10 AM
> To: amibroker@xxxxxxxxxxxxxxx
> Subject: [amibroker] Re: Are your Composites accurate???
>
>
> Herman,
> when we scan a database, Amibroker "sees" the existing data only.
> If for ADLAC the 15/4 exists and all the rest data are missing, scan
> will not see an empty value for 16/4. It will just see nothing, like
> a Sunday. So, if I understand well, the open==empty is useless.
> Any composite ticker graph will stop at 15/4, when you select ADLAC.
> We need a positive fact to count, this is why I thought to assign an
> 1, if the stock is present.
> Correct me if I am wrong.
> Dimitris Tsokakis
> --- In amibroker@xxxx, "Herman van den Bergen" <psytek@xxxx> wrote:
> > Hi all,
> >
> > The number of historical bars for each stock in your data base may
> vary. For
> > example if I check the number of bars for my N100 data they vary
> from 612 to
> > 2597.
> >
> > Depending on the type of composites you create this may be an
> important
> > issue to consider. If you want to average a parameter for the N100,
> for
> > example, you cannot simply divide each bar in your composite by
> 100; unless
> > you have checked that data is present for ALL 100 stocks, for each
> bar, for
> > the period sampled.
> >
> > I use this preliminary Scan to give me the numbers I need to
> calculate
> > accurate averages. See the attached capture for my N100.
> >
> > EMPTY = -1^10;
> > AddToComposite(IIf(Open == EMPTY,0,1),"_DataPresent","x",3);
> > Buy= 0;
> > Filter=1;
> >
> > Can this be improved?
> > Note in my capture that data can have holes and that you have to
> consider
> > how to handle those.
> >
> > best regards,
> > Herman.
>
>
>
>
>
> Your use of Yahoo! Groups is subject to http://docs.yahoo.com/info/terms/
>
>
Attachment:
jpg00105.jpg
Attachment:
Description: "Description: JPEG image"
|