|
hi Fred,
thanks for your nice words. I do not see the
need for a change to LPeriod-1. In the code below I add a parameter
"shift". When you shift it in the parameter window you see what is line is
actually calculated, where it is located and where it puts the resulting
bandwidth number (see also interpretation window). It seems OK to me. Especially
when you use C = Cum(1) you can see that the least squares fit lies
exactly on to of the Cum(1) function.
Your second point I'm not sure,
rgds, Ed
SetBarsRequired(10000,10000);
shift = Param( "shift", 0, 0, 100, 1 );
PPeriod = Param( "P Period", 12,
2, 100, 1 );
LPeriod = Param( "L Period", 12,
2, 100, 1 );
NPeriod = Param( "N Period", 2,
0, 10, 0.25 );
Bandwidth = Null;
//C =
Cum(1);
EmaP = EMA(C,PPeriod);
saveLine
= Null;
aaa = LinRegIntercept(C, LPeriod);
bbb = LinRegSlope(C, LPeriod);
for (i = BarCount - 1 - shift; i < BarCount - shift; i++) {
AE =
0;
for (j = 0; j < LPeriod; j++) {
// calculate
point on regression line
RegPoint =
aaa[ i ] + bbb[ i ] * j;
//
calculate the standard error using the distance from
line
AE = AE + (C[ (i - LPeriod) + j +
1 ] - RegPoint) ^
2;
printf( "Regpoint: %g\n", Regpoint );
printf( "C[
(i - LPeriod) + j + 1 ]: %g\n",
C[ (i - LPeriod) +
j + 1 ] );
saveLine[ i - LPeriod + j +
1 ] = RegPoint;
}
//
standard error
avErr = (AE / LPeriod) ^ 0.5;
// save the
bandwidth
Bandwidth[ i ] = avErr * NPeriod;
}
SetChartOptions(0, chartShowDates);
GraphXSpace = 5;
Plot(saveLine,"",colorRed,1);
Plot(saveLine,"",colorYellow,9);
Plot(C,"",colorWhite,64);
Plot(saveLine,"",colorRed,1);
EmaP = IIf(Bandwidth,EmaP,Null);
PlotShapes(IIf(Bandwidth,shapeDownTriangle,0),colorGold, layer = 0, yposition = EmaP + Bandwidth, offset
= 0 );
PlotShapes(IIf(Bandwidth,shapeUpTriangle,0),colorGold, layer = 0, yposition = EmaP - Bandwidth, offset
= 0 );
"";
"";
Title=Name()+ ",
O: "+WriteVal(O)+
", H: "+WriteVal(H)+ ", L: "+WriteVal(L)+ ", C: "+WriteVal(C) +
----- Original Message -----
Sent: Monday, March 06, 2006 11:15
AM
Subject: Re: [amibroker] Help with
Kirshenbaum indicator
With much interest i read the contributions
regarding the Kirshenbaum indicator. Special thanks to Ed (Emp62) his
contributions gave me a valuable insight how loops actually
work...
When thorougly checking the calculated values, i
came across different values (LinReg values)
In my opinion the formula should be slightly
changed as follows:
Original Empc posting:
aaa = LinRegIntercept(C, LPeriod);
bbb = LinRegSlope(C, LPeriod);
in my
opinion should be changed to:
aaa = LinRegIntercept(C, LPeriod-1);
bbb = LinRegSlope(C, LPeriod-1);
Another thought when
studying the indicator:
The "standard error calculation" measures
the distance (C-Regpoint) for Lperiod days.
Obviously it make a huge difference whether the symbol is a
10$ trading stock or the Dow Index of 11000 points.
So my suggestion is to normalize the
calculation for instance (C/Regpoint-1)*100 to calculate the procentual
change.
Fred
SetBarsRequired(10000,10000);
PPeriod = Param( "P Period", 12, 2, 100, 1 );
LPeriod = Param( "L Period", 12, 2, 100, 1 );
NPeriod = Param( "N Period", 2, 0, 10, 0.25 );
Bandwidth = 0;
EmaP = EMA(C,PPeriod);
aaa = LinRegIntercept(C, LPeriod);
bbb = LinRegSlope(C, LPeriod);
for (i = LPeriod; i <
BarCount; i++) {
AE =
0;
for (j = 0; j < LPeriod; j++) {
// calculate point on regression
line
RegPoint = aaa[ i ] +
bbb[ i ] * j;
//
calculate the standard error using the distance from
line
AE = AE +
(C[ (i -
LPeriod) + j + 1 ] -
RegPoint) ^ 2;
}
// standard error
avErr = (AE /
LPeriod) ^ 0.5;
// save the bandwidth
Bandwidth[ i ] =
avErr * NPeriod;
}
Buy = Ref(Cross(EmaP - Bandwidth,C),-1); BuyPrice = O;
Sell = Ref(Cross(C, EmaP +
Bandwidth),-1);
SellPrice =
O;
SetChartOptions(0,
chartShowDates);
GraphXSpace = 5;
Plot(C,"",colorWhite,64);
Plot(EmaP,"",colorWhite,1);
Plot(EmaP + Bandwidth,"",colorRed,1);
Plot(EmaP - Bandwidth,"",colorBrightGreen,1);
PlotShapes(IIf(Buy,shapeUpTriangle,0),colorWhite, layer =
0, yposition =
BuyPrice,
offset = 0 );
PlotShapes(IIf(Sell,shapeDownTriangle,0),colorYellow, layer = 0, yposition = SellPrice, offset =
0 );
----- Original Message -----
Sent: Saturday, March 04, 2006 11:14
PM
Subject: RE: [amibroker] Help with
Kirshenbaum indicator
With
those settings I do not see there being a buy or sell on the chart since
the close price is not out of the bands
anywhere
Am I
missing something?
Mark
I changed 1 line in the code.
I think it is good now. See chart below which resembles your original
chart very good (settings I use for that chart are: 5, 7,
4.5)
SetBarsRequired(10000,10000);
PPeriod = Param(
"P
Period",
12,
2,
100,
1 );
LPeriod = Param(
"L
Period",
12,
2,
100,
1 );
NPeriod = Param(
"N
Period",
2,
0,
10,
0.25 );
Bandwidth = 0;
EmaP = EMA(C,PPeriod);
aaa =
LinRegIntercept(C, LPeriod);
bbb =
LinRegSlope(C, LPeriod);
for (i =
LPeriod; i < BarCount;
i++) {
AE =
0;
for (j =
0; j <
LPeriod; j++) {
//
calculate point on regression line
RegPoint = aaa[ i ] + bbb[ i ] *
j;
//
calculate the standard error using the distance from
line
AE = AE + (C[ (i - LPeriod) + j +
1 ] -
RegPoint) ^ 2;
}
//
standard error
avErr = (AE / LPeriod) ^ 0.5;
// save
the bandwidth
Bandwidth[ i ] = avErr * NPeriod;
}
SetChartOptions(0,
chartShowDates);
GraphXSpace =
5;
Plot(C,"",colorWhite,64);
Plot(EmaP,"",colorWhite,1);
Plot(EmaP +
Bandwidth,"",colorRed,1);
Plot(EmaP -
Bandwidth,"",colorBrightGreen,1);
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
|