[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

RE: Another calculation question



PureBytes Links

Trading Reference Links

Martin,

I can't blame MS on this one.  When I was moving stuff around, I made a
typo.  It was totally my fault.  I had tested each component and it worked.
Then I moved stuff around and didn't retest everyone.  I made a type using
your CUM() technique (which is working fine, BTW).

I did have some problems with MS and its precision but was able to work
around that a long time ago (MS for DOS).

Guy

Never be afraid to try something new. Remember, amateurs built the ark,
professionals built the Titanic.

-----Original Message-----
From: owner-metastock@xxxxxxxxxxxxx [mailto:owner-metastock@xxxxxxxxxxxxx]On
Behalf Of Martin Haesler
Sent: Sunday, August 20, 2000 12:21 AM
To: metastock@xxxxxxxxxxxxx
Subject: Re: Another calculation question

Guy

I'll pass on an area where I had a problem with Metastock's eod
capabilities.

When I created Don Fishers DGL indicator, I decided to use a date format of
YYYYMMDD.

I then extracted the year and month by division.

The remainder (after dividing by 1000000) should have been the "days" DD.
However, this number turned out to be very unreliable indicating the
accuracy of MS arithmetic was limited. When I changed the format to YYMMDD
reducing the division to 10000  all worked well. This of course mean I had
to allow for the year 2000 in a different way but that's another story.

The point I wanted to bring home to you, is that you cannot assume that MS
will calculate anywhere near accurately numbers to more than 6 significant
digits.

With your current situation, I would bring across each of your variables
into a test indicator and plot it, verifying it is the value you expect.
Then do your addition and go on from there.

Regards ... Martin


----- Original Message -----
From: "Guy Tann" <grt@xxxxxxxxxxxx>
To: "Metastock User Group" <metastock-list@xxxxxxxxxxxxx>
Sent: Sunday, August 20, 2000 4:00 PM
Subject: Another calculation question


> All and Ton, Martin, et al:
>
> Well, I'm moving along with my programming thanks to Ton, Martin and all
> other contributors, but I've reached an impasse trying to get to one of my
> more difficult calculations.  I am debugging each one of my calculations
and
> plotting its individual result in MS and comparing it to a debugging
program
> I've written in Clipper.  I feel that if I debug each individual
> calculation, when I get to the end, everything should work.
>
> Well, that worked for while, until I reached the following calculation.
> Something about the best laid plans of mice and men. :)
>
> COMHACR:= ( FmlVar("COMH-BASICS","COMHCRR") +
> FmlVar("COMH-BASICS","COMHCYY") +  FmlVar("COMH-BASICS","COMHCY4") +
> FmlVar("COMH-BASICS","COMHCR4") +  FmlVar("COMH-BASICS","COMHCR7") +
> COMHCY7 ) / 3;
> COMHACR;
>
> I have previously debugged COMHCRR, COMHCYY, COMHCY4, COMHCR4, COMHCR7,
and
> COMHCY7 and they all calculate properly and will plot if I set it up.
This
> calculation simply adds the six values and divides the total by 3.
Nothing
> sophisticated at all.  I then moved this variable calculation to another
> blank indicator to try it.  It still didn't work, so I then tried to
comment
> out everything but one of the components and that didn't work either
> (COMHCRR) but when I go back to that variable and set it up to plot, it
> works fine.  I'm terminally confused here.
>
> The only thing I can think is that I've exceeded MS programming
> capabilities, but you would think I would at least get an error message.
I
> mean we're not talking brain surgery here, so I assume that I'm doing
> something stupid.  :) For instance, when I reached the point that I had
> exceeded the calculations one indicator can support (something about
binary
> but I didn't write it down) I moved a bunch of the calculations to another
> indicator.
>
> Is there a limit on the number of indicators that you can program in MS?
I
> wouldn't think I'm near any limit, but???  I have a bunch of old
> calculations and system testers out there.  I hadn't planned on it, but
> should I go out and delete them to make room for the newer stuff?
>
> TIA,
>
> Guy
>
> Never be afraid to try something new. Remember, amateurs built the ark,
> professionals built the Titanic.
>
>