|
PureBytes Links
Trading Reference Links
|
Fred, it took me several days to work on/off on this. Your
Cycleconstant definition without loops: brilliant! So array+number
results in new array where number is added to each array element.
Cool, I did not know that.
I have worked to define StDevCum for individual bars through
recursion which was a painful but useful exercise in itself:
// function StDevCum recursion-style:
array = C ; // set array here
mean[0] = array[0] ;
StDevcum[0] = 0 ;
for (i = 1 ; i < BarCount ; i++ )
{
mean[i] = (i/(i+1))*mean[i-1] + (array[i])/(i+1) ;
StDevCum[i] = sqrt( (i/(i+1))*(StDevCum[i-1])^2 + (1/i)*(array[i] -
mean[i])^2 ) ;
}
//
But blending this into the larger code does not improve the speed.
This is what I came up with:
// code start
/* NOTE: the first few lines below are IRrelevant, only to create
random Cycle values between 10 and 50 and some Oscillator function,
in order for the lower part to work */
function Randomize(a,b)
{ return Random(1)*(b-a)+a ; }
Cycle = int( Randomize(10,50) ) ;
function Oscillator(n)
{ return Randomize(-50,n) ; }
/* Here starts the relevant part */
//SetBarsRequired( 100000, 100000 );
function Cycleconstant(number)
{ return Cum(0) + Cycle[ number ] ; }
for (i = BarCount - 1 ; i > 0 ; i-- )
{
array = Oscillator( Cycleconstant(i) ) ;
mean[0] = array[0] ;
StDevcumOfOscillator[0] = 0 ;
for (j = 1 ; j <= i ; j++ )
{
mean[j] = (j/(j+1)) * mean[j-1] + (array[j])/(j+1) ;
StDevCumOfOscillator[j] = sqrt( (j/(j+1))*(StDevCumOfOscillator[j-1])
^2 + (1/j)*(array[j] - mean[j])^2 ) ;
}
}
Plot(0,"",colorBlack);
Plot(Oscillator(Cycle),"Oscillator(Cycle)",colorBlue);
Plot(StDevCumOfOscillator,"StDevCumOfOscillator",colorGreen);
GraphZOrder = 1;
// code end
Although this gives the correct result it actually takes a bit longer
to execute. Probably I misinterpreted your suggestion ("you can
write
the full calc for indivdual bars w/o loops as well") because
although
this works on individual bars it contains a loop with a sub-loop
here. But how to do without ??
And then i (in the first loop) I have to let run from high to low
otherwise previously calculated StDevCumOfOscillator array elements
are overridden. Oeeff..... I find this very complex stuff, sure way
to a quick headache.....
If you could give any additional advice this will be gratefully
received. If not, thanks very much for your help so far.
-treliff
--- In amibroker@xxxxxxxxxxxxxxx, "Fred" <ftonetti@xxxx> wrote:
> In any case ... writing loops for stuff is leftover programmer
> ability if you will ... I suffered from the same thing for awhile
> back when I first started using AB ...
>
> For example ...
>
> Your ...
>
> function CycleConstant(number)
> {
> for ( i = 0 ; i < BarCount ; i++ )
> {
> result[ i ] = Cycle[ number ] ;
> }
> return result;
> }
>
> Can be written ...
>
> Result = Cum(0) + Cycle[number];
>
> This at least takes the lowest level loop out of play ...
>
> When you get around to writing the standard dev part you can write
> the full calc for indivdual bars w/o loops as well ... so there
> should only be one loop necessary ...
>
> --- In amibroker@xxxxxxxxxxxxxxx, "treliff" <treliff@xxxx> wrote:
> > Fred, ALL bars indeed. This is not short-term-TA but long-term-
> > statistics (an attempt at least :-). 3 years of daily bars is
> > minimum, 10 years would be better.
> >
> > Your amazement (Is that really what you are wanting ?) though may
> > have pointed me in the right direction: perhaps I should de-link
> > these long term standevs from the code. I am thinking about
> > artificial AddToComposite tickers to store these measurements. No
> > real "Composites" but multiple per-stock tickers like
> >
> > Ticker+"Cycle"+"10to13"
> > Ticker+"Cycle"+"14to17" etc
> >
> > (certainly incorrect code, I'd have to work this out)
> >
> > That would be about 10 ATC tickers per stock (40 different cycles
> > with 4 storage spaces OHLC for each ticker). One time-consuming
ATC
> > run over multiple stocks.
> >
> > Remains to be seen though if the code finds accessing data in
> > truckloads of ATC tickers much more pleasant than doing the
> > calculations itself.....
> >
> > Thanks for thinking along.
> >
> > -treliff
> >
> > --- In amibroker@xxxxxxxxxxxxxxx, "Fred" <ftonetti@xxxx> wrote:
> > > Is that really what you are wanting ?
> > >
> > > Standard deviation of all numbers from the first bar through
the
> > > current bar ? or at the current bar are you wanting standard
> > > deviation of the numbers of the most current n bars where n is
> the
> > > current cycle ?
> > >
> > > --- In amibroker@xxxxxxxxxxxxxxx, "treliff" <treliff@xxxx>
wrote:
> > > > Fred, thanks for your comments.
> > > >
> > > > Correct about standard deviation, the defined StDevCum(array)
> is
> > > > really StDev(array, BarIndex() + 1) but this one does not
work.
> > > >
> > > > However I'm not sure this is what you are referring to
because
> > this
> > > > is not really a "loop".
> > > >
> > > > Could you perhaps be more specific, maybe point me in the
> > direction
> > > > of a loop-less re-write? Highly appreciated.
> > > >
> > > > -treliff
> > > >
> > > > --- In amibroker@xxxxxxxxxxxxxxx, "Fred" <ftonetti@xxxx>
wrote:
> > > > > Regardless ... It would appear the only loop that's
required
> is
> > > the
> > > > > one that calc's standard deviation as the built in AFL
> version
> > > does
> > > > > not take an array for length. The rest of the loops can be
> > > > replaced
> > > > > by arrays and in some cases totally done away with.
> > > > >
> > > > > --- In amibroker@xxxxxxxxxxxxxxx, "Terry" <MagicTH@xxxx>
> wrote:
> > > > > > No, I'm not sure, but this has been discussed before. The
> > last
> > > I
> > > > > recall
> > > > > > Tomasz said not to use this anymore...but we're counting
on
> > > > > my "recall"
> > > > > > so that's not reliable ;-)
> > > > > >
> > > > > > I do know that AB will use as many bars as necessary to
get
> > > > > the "right"
> > > > > > answer. It's very smart in that regard. Why don't you
> record
> > > some
> > > > > > results, take out the SetBarsRequired and see if the
> answers
> > > are
> > > > > the
> > > > > > same and if the speed is improved?
> > > > > > --
> > > > > > Terry
> > > > > >
> > > > > > | -----Original Message-----
> > > > > > | From: amibroker@xxxxxxxxxxxxxxx
> > > > > [mailto:amibroker@xxxxxxxxxxxxxxx] On
> > > > > > | Behalf Of treliff
> > > > > > | Sent: Tuesday, August 23, 2005 09:04
> > > > > > | To: amibroker@xxxxxxxxxxxxxxx
> > > > > > | Subject: [amibroker] Re: some looping help
needed .......
> > > > > > |
> > > > > > | Terry, no I don't need 100,000 bars (currently testing
on
> > > > several
> > > > > > | years of daily bars) but I do want the calculation
> (standard
> > > > > > | deviation) over all bars, so back to bar zero from any
> > > starting
> > > > > bar.
> > > > > > |
> > > > > > | While AB Help menu says that SetBarsRequired is only
> > necessary
> > > > > > | outside pure AFL (having to do with "QuickAFL") I find
> many
> > > > pure
> > > > > AFL
> > > > > > | code examples that contain looping having this
> > > > > > |
> > > > > > | setbarsrequired( 100000, 100000 ) // require all past
and
> > all
> > > > > future
> > > > > > | bars
> > > > > > |
> > > > > > | statement at the beginning (like in AFL library or in
> this
> > > > > forum).
> > > > > > |
> > > > > > | Are you sure it can be left out?
> > > > > > |
> > > > > > | I appreciate your help. Thanks.
> > > > > > |
> > > > > > | -treliff
> > > > > > |
> > > > > > | --- In amibroker@xxxxxxxxxxxxxxx, "Terry"
<MagicTH@xxxx>
> > > wrote:
> > > > > > | > Do you really need 100,000 bars to get the right
answer?
> > > > > > | > Seems like not forcing this many bars would speed
> things
> > up
> > > a
> > > > > lot.
> > > > > > | > --
> > > > > > | > Terry
> > > > > > | > | -----Original Message-----
> > > > > > | > | From: amibroker@xxxxxxxxxxxxxxx
> > > > > > | [mailto:amibroker@xxxxxxxxxxxxxxx] On
> > > > > > | > | Behalf Of treliff
> > > > > > | > | Sent: Monday, August 22, 2005 20:10
> > > > > > | > | To: amibroker@xxxxxxxxxxxxxxx
> > > > > > | > | Subject: [amibroker] some looping help
needed .......
> > > > > > | > |
> > > > > > | > | Hoping for some help from Expert Coders.
> > > > > > | > |
> > > > > > | > | Code in question is part of a larger code and
creates
> an
> > > > > > | incredible
> > > > > > | > | drag, slow chart scrolling etc., understandably so
in
> > > view
> > > > > of the
> > > > > > | > | amount of procedures. Hope this can be
> > > simplified/improved.
> > > > > > | > | Appreciate anyone's time to take a look and
possibly
> > help
> > > > me
> > > > > out.
> > > > > > | > |
> > > > > > | > | I have an array called "Cycle" that contains
integers
> > > > between
> > > > > > | > | 10 and
> > > > > > | > | 50.
> > > > > > | > |
> > > > > > | > | I have a function called "Oscillator" that, working
> on
> > any
> > > > > > | > | Cycle
> > > > > > | > | array, returns some oscillator, meaning it is
limited
> > on
> > > up-
> > > >
> > > > > and
> > > > > > | > | downside and has Mean approximately zero (like an
> > > irregular
> > > > > > | Sinus).
> > > > > > | > |
> > > > > > | > | My challenge is that once Cycle has been generated
> > > (earlier
> > > > > in my
> > > > > > | > | code) I want to calculate the standard deviation of
> > > > > Oscillator
> > > > > > | over
> > > > > > | > | all previous bars using strictly the Cycle value of
> that
> > > > > > | particular
> > > > > > | > | Bar (not the Cycle array which has different
values).
> > > > > > | > |
> > > > > > | > | For example:
> > > > > > | > | - Bar 300 has cycle value 27
> > > > > > | > | - I create an array with value "27" in all array
> > elements
> > > > > > | > | - then calculate Oscillator of this "27" array
> > > > > > | > | - calculate the standard deviation of all
Oscillator
> > > values
> > > > > up to
> > > > > > | Bar
> > > > > > | > | 300
> > > > > > | > | - place this particular value in Bar 300 of a new
> array
> > > > > > | > | - execute this procedure for all Bars separately
> > > > > > | > |
> > > > > > | > | Below is the code I made. TIA for any advice.
> > > > > > | > |
> > > > > > | > | // code start
> > > > > > | > |
> > > > > > | > | /* NOTE: the first few lines below are IRrelevant,
> only
> > > to
> > > > > create
> > > > > > | > | random Cycle values between 10 and 50 and some
> > Oscillator
> > > > > > | function,
> > > > > > | > | in order for the lower part to work */
> > > > > > | > |
> > > > > > | > | function Randomize(a,b)
> > > > > > | > | { return Random(1)*(b-a)+a ; }
> > > > > > | > |
> > > > > > | > | Cycle = int( Randomize(10,50) ) ;
> > > > > > | > |
> > > > > > | > | function Oscillator(n)
> > > > > > | > | { return Randomize(-50,n) ; }
> > > > > > | > |
> > > > > > | > | /* HERE starts the relevant part */
> > > > > > | > |
> > > > > > | > | SetBarsRequired( 100000, 100000 );
> > > > > > | > |
> > > > > > | > | function StDevCum(array)
> > > > > > | > | { return sqrt( ( (BarIndex()+1) * Cum(array^2) -
(Cum
> > > > (array))
> > > > > > | ^2 ) /
> > > > > > | > | (BarIndex()+1)^2 ); }
> > > > > > | > |
> > > > > > | > | /* StDevCum calculates standard deviation of array
> from
> > > Bar
> > > > > zero
> > > > > > | up
> > > > > > | > | to current Bar. */
> > > > > > | > |
> > > > > > | > | function Cycleconstant(number)
> > > > > > | > | { for ( i = 0 ; i < BarCount ; i++ )
> > > > > > | > | { result[ i ] = Cycle[ number ] ; }
> > > > > > | > | return result; }
> > > > > > | > |
> > > > > > | > | /* Cycleconstant fills a complete array with a
Cycle
> > > array
> > > > > element
> > > > > > | > | value (number). */
> > > > > > | > |
> > > > > > | > | for (j = 0 ; j < BarCount ; j++ )
> > > > > > | > | { y = StDevCum( Oscillator( Cycleconstant( j ) ) ) ;
> > > > > > | > | StDevCumOfOscillator[ j ] = y[ j ] ; }
> > > > > > | > |
> > > > > > | > | /* for each separate Bar this first works
Oscillator
> on
> > > > > > | Cycleconstant
> > > > > > | > | array of that Bar, then calculates standard
deviation
> > > (from
> > > > > Bar 0)
> > > > > > | > | and places result in Bar value (array element) of
new
> > > array
> > > > > > | > | StDevCumOfOscillator. */
> > > > > > | > |
> > > > > > | > | Plot(0,"",colorBlack);
> > > > > > | > | Plot(Oscillator(Cycle),"Oscillator
(Cycle)",colorBlue);
> > > > > > | > | Plot
> > > (StDevCumOfOscillator,"StDevCumOfOscillator",colorRed);
> > > > > > | > | GraphZOrder = 1;
> > > > > > | > |
> > > > > > | > | // code end
> > > > > > | > |
> > > > > > | > |
> > > > > > | > |
> > > > > > | > |
> > > > > > | > |
> > > > > > | > | ------------------------ Yahoo! Groups Sponsor -----
--
> --
> > --
> > > --
> > > > -
> > > > > -----
> > > > > > | -~--
> > > > > > | > | >
> > > > > > | > | <font face=arial size=-1><a
> > > > > > | > |
> > > > > > |
> > > > >
> > > >
> > >
> >
>
href="http://us.ard.yahoo.com/SIG=12h0dd89q/M=362131.6882500.7825259.
> > > > > 1
> > > > > > | > |
> > > > > > |
> > > > >
> > > >
> > >
> >
>
493532/D=groups/S=1705632198:TM/Y=YAHOO/EXP=1124770234/A=2889190/R=0/
> > > > > S
> > > > > > | > | IG=10r90krvo/*http://www.thebeehive.org
> > > > > > | > | ">Put more honey in your pocket. (money matters
made
> > easy)
> > > > > > | Welcome to
> > > > > > | > | the Sweet Life - brought to you by One
> > Economy</a>.</font>
> > > > > > | > | ----------------------------------------------------
--
> --
> > --
> > > --
> > > > -
> > > > > -----
> > > > > > | --~-
> > > > > > | > | >
> > > > > > | > |
> > > > > > | > | Please note that this group is for discussion
between
> > > users
> > > > > only.
> > > > > > | > |
> > > > > > | > | To get support from AmiBroker please send an e-mail
> > > > directly
> > > > > to
> > > > > > | > | SUPPORT {at} amibroker.com
> > > > > > | > |
> > > > > > | > | For other support material please check also:
> > > > > > | > | http://www.amibroker.com/support.html
> > > > > > | > |
> > > > > > | > |
> > > > > > | > | Yahoo! Groups Links
> > > > > > | > |
> > > > > > | > |
> > > > > > | > |
> > > > > > | > |
> > > > > > | > |
> > > > > > |
> > > > > > |
> > > > > > |
> > > > > > |
> > > > > > |
> > > > > > | ------------------------ Yahoo! Groups Sponsor ---------
--
> --
> > --
> > > --
> > > > -
> > > > > --~--
> > > > > > | >
> > > > > > | <font face=arial size=-1><a
> > > > > > |
> > > > >
> > > >
> > >
> >
>
href="http://us.ard.yahoo.com/SIG=12hhmiq6u/M=362131.6882500.7825259.
> > > > > 1
> > > > > > |
> > > > >
> > > >
> > >
> >
>
493532/D=groups/S=1705632198:TM/Y=YAHOO/EXP=1124816731/A=2889190/R=0/
> > > > > S
> > > > > > | IG=10r90krvo/*http://www.thebeehive.org
> > > > > > | ">Put more honey in your pocket. (money matters made
> easy)
> > > > > Welcome to
> > > > > > | the Sweet Life - brought to you by One
Economy</a>.</font>
> > > > > > | --------------------------------------------------------
--
> --
> > --
> > > --
> > > > -
> > > > > ---~-
> > > > > > | >
> > > > > > |
> > > > > > | Please note that this group is for discussion between
> users
> > > > only.
> > > > > > |
> > > > > > | To get support from AmiBroker please send an e-mail
> > directly
> > > to
> > > > > > | SUPPORT {at} amibroker.com
> > > > > > |
> > > > > > | For other support material please check also:
> > > > > > | http://www.amibroker.com/support.html
> > > > > > |
> > > > > > |
> > > > > > | Yahoo! Groups Links
> > > > > > |
> > > > > > |
> > > > > > |
> > > > > > |
------------------------ Yahoo! Groups Sponsor --------------------~-->
Get fast access to your favorite Yahoo! Groups. Make Yahoo! your home page
http://us.click.yahoo.com/dpRU5A/wUILAA/yQLSAA/GHeqlB/TM
--------------------------------------------------------------------~->
Please note that this group is for discussion between users only.
To get support from AmiBroker please send an e-mail directly to
SUPPORT {at} amibroker.com
For other support material please check also:
http://www.amibroker.com/support.html
Yahoo! Groups Links
<*> To visit your group on the web, go to:
http://groups.yahoo.com/group/amibroker/
<*> To unsubscribe from this group, send an email to:
amibroker-unsubscribe@xxxxxxxxxxxxxxx
<*> Your use of Yahoo! Groups is subject to:
http://docs.yahoo.com/info/terms/
|