|
PureBytes Links
Trading Reference Links
|
While John is trying to fix his floating point error, I'm trying to fix
a Divide-by-zero error message when adding "Standard Error Bands"
indicator to a TS2ki chart.
Anyone describing the problem and fix is Hero at my house. (Until my
wife gets back)
Here's the Stocks & Comod Mag's version which also triggers the same
error message as the version supplied with TS2ki:
TRADESTATION
To create the standard error bands described by Jon Andersen in
"Standard error bands" in the September 1996 STOCKS & COMMODITIES, you
must calculate the beta and alpha coefficients of the linear regression.
First, create the calcB and calcA user functions, which refer to the
beta and alpha coefficients. Be sure to create and verify these
functions before building the indicator. Create calcB first, then calcA.
These functions are used in calculating the regression coefficients for
the standard error bands.
Type: User Function
Name: calcB
Inputs: Length(Numeric);
Vars: X(0);
X = BarNumber;
Value1 = Summation(X * Close, Length) - (Length * Average(X, Length)
*Average(Close, Length));
Value2 = Summation(Square(X), Length) - (Length * Square(Average(X,
Length)));
calcB = Value1/Value2;
Type: User Function
Name: calcA
Inputs: Length(Numeric);
Vars: X(0);
X = BarNumber;
calcA = Average(Close, Length) - (calcB(Length) * Average(X, Length));
Next, build the standard error bands indicator. The "length" value
specifies the period of the linear regression line; the default is 21.
The SDeg value specifies the smoothing factor that is used to smooth the
linear regression and standard error bands; the default is 3.
Type: Indicator
Name: Std Error Bands Smooth
Inputs: Length(21), SDeg(3);
Vars: LinRegY(0), X(0), StdErr(0), LinRegS(0), SErr(0);
LinRegY = LinearRegValue(Close, Length, 0);
X = CurrentBar;
IF CurrentBar > Length then Begin
Value1 = (Summation(Square(Close), Length)) - ((calcA(Length) *
Summation(Close, Length))) - ((calcB(Length) *
Summation(X * Close, Length)));
Value2 = Length - 2;
StdErr = SquareRoot(Value1 / Value2);
LinRegS = Average(LinRegY, SDeg);
SErr = 2 * Average(StdErr, SDeg);
Plot1(LinRegS, "LinRegS");
Plot2(LinRegS + SErr, "+StdErrS");
Plot3(LinRegS - SErr, "-StdErrS");
End;
The indicator for the related %A standard error bands is built
using the code below. It's best plotted with the bar type set to
"histogram."
Type: Indicator
Name: Standard Error Bands %A
Inputs: Length(21), SDeg(3);
Vars: LinRegY(0), X(0), StdErr(0), LinRegS(0), SErr(0), PcntA(0);
LinRegY = LinearRegValue(Close, Length, 0);
X = CurrentBar;
IF CurrentBar > Length then Begin
Value1 = (Summation(Square(Close), Length)) - ((calcA(Length) *
Summation(Close, Length))) - ((calcB(Length) *
Summation(X * Close, Length)));
Value2 = Length - 2;
StdErr = SquareRoot(Value1 / Value2);
LinRegS = Average(LinRegY, SDeg);
SErr = 2 * Average(StdErr, SDeg);
PcntA = (Close - (LinRegS-SErr)) / ((LinRegS+SErr)-(LinRegS-SErr))
* 100;
Plot1(PcntA, "%A");
End;
This code is available at both the Omega Research Forum on America
On-Line as well as at Omega's Web site.
-- Gaston Sanchez, Omega Research
Internet: http://www.omegaresearch.com
TRADESTATION
To create the standard error bands described by Jon Andersen in "Standard error bands" in the September 1996 STOCKS & COMMODITIES, you must calculate the beta and alpha coefficients of the linear regression. First, create the calcB and calcA user functions, which refer to the beta and alpha coefficients. Be sure to create and verify these functions before building the indicator. Create calcB first, then calcA. These functions are used in calculating the regression coefficients for the standard error bands.
Type: User Function
Name: calcB
Inputs: Length(Numeric);
Vars: X(0);
X = BarNumber;
Value1 = Summation(X * Close, Length) - (Length * Average(X, Length) *Average(Close, Length));
Value2 = Summation(Square(X), Length) - (Length * Square(Average(X, Length)));
calcB = Value1/Value2;
Type: User Function
Name: calcA
Inputs: Length(Numeric);
Vars: X(0);
X = BarNumber;
calcA = Average(Close, Length) - (calcB(Length) * Average(X, Length));
Next, build the standard error bands indicator. The "length" value specifies the period of the linear regression line; the default is 21. The SDeg value specifies the smoothing factor that is used to smooth the linear regression and standard error bands; the default is 3.
Type: Indicator
Name: Std Error Bands Smooth
Inputs: Length(21), SDeg(3);
Vars: LinRegY(0), X(0), StdErr(0), LinRegS(0), SErr(0);
LinRegY = LinearRegValue(Close, Length, 0);
X = CurrentBar;
IF CurrentBar > Length then Begin
Value1 = (Summation(Square(Close), Length)) - ((calcA(Length) *
Summation(Close, Length))) - ((calcB(Length) *
Summation(X * Close, Length)));
Value2 = Length - 2;
StdErr = SquareRoot(Value1 / Value2);
LinRegS = Average(LinRegY, SDeg);
SErr = 2 * Average(StdErr, SDeg);
Plot1(LinRegS, "LinRegS");
Plot2(LinRegS + SErr, "+StdErrS");
Plot3(LinRegS - SErr, "-StdErrS");
End;
The indicator for the related %A standard error bands is built using the code below. It's best plotted with the bar type set to "histogram."
Type: Indicator
Name: Standard Error Bands %A
Inputs: Length(21), SDeg(3);
Vars: LinRegY(0), X(0), StdErr(0), LinRegS(0), SErr(0), PcntA(0);
LinRegY = LinearRegValue(Close, Length, 0);
X = CurrentBar;
IF CurrentBar > Length then Begin
Value1 = (Summation(Square(Close), Length)) - ((calcA(Length) *
Summation(Close, Length))) - ((calcB(Length) *
Summation(X * Close, Length)));
Value2 = Length - 2;
StdErr = SquareRoot(Value1 / Value2);
LinRegS = Average(LinRegY, SDeg);
SErr = 2 * Average(StdErr, SDeg);
PcntA = (Close - (LinRegS-SErr)) / ((LinRegS+SErr)-(LinRegS-SErr)) * 100;
Plot1(PcntA, "%A");
End;
This code is available at both the Omega Research Forum on America On-Line as well as at Omega's Web site.
-- Gaston Sanchez, Omega Research
Internet: http://www.omegaresearch.com
|