The r-squared calculation that Jack Karczewski focuses on in "Identifying market trends" is very useful in identifying trending and nontrending periods in the markets. This month, I'll present a revised version of the r-squared indicator given in Jon Andersen's September 1996 STOCKS & COMMODITIES article, "Standard error bands." I'll also include two PaintBar studies that identify trending and nontrending periods by painting the price bars. All three studies reference a function named "coeffR" for their basic calculation.
The "coeffR" function must be created and verified in the Power Editor prior to creating the studies. The EasyLanguage code for the function follows:
Type: Function
Name: coeffR
Inputs: Length(Numeric);
Vars: X(0), UpEQ(0), LowEQ1(0), LowEQ2(0), LowEQT(0);
X = BarNumber;
UpEQ = Summation(X * Close, Length) - (Length * Average(X, Length) * Average(Close, Length));
LowEQ1 = Summation(Square(X), Length) - (Length * Square(Average(X, Length)));
LowEQ2 = Summation(Square(Close), Length) - (Length * Square(Average(Close, Length)));
IF LowEQ1 * LowEQ2 > 0 Then
LowEQT = SquareRoot(LowEQ1 * LowEQ2);
IF LowEQT <> 0 Then
CoeffR = UpEQ / LowEQT;
The next step is to develop the r-squared indicator. The r-squared indicator plots the r-squared line as well as a zone for the trending and nontrending levels. In addition, a dot will be displayed on the r-squared line when it has been above/below a trending/nontrending level for a specified number of bars. The Length input determines the length of the r-squared calculation. The Smooth input allows for the r-squared line to be smoothed with a moving average. The Trend input determines the level above which the market is considered trending. The NoTrend input determines the level below which the market is considered nontrending. The RunLen input determines how many consecutive bars must be above/below the trending/nontrending levels in order for a mark to be placed on the r-squared line and an alert to be triggered. The suggested format settings are displayed after the EasyLanguage code.
Type: Indicator
Name: R-Squared
Inputs: Length(30), Smooth(1), Trend(.7), NoTrend(.3), RunLen(10);
Vars: R(0), SquareR(0), AvgSqrR(0);
R = coeffR(Length);
SquareR = Square(R);
IF Smooth > 0 Then
AvgSqrR = Average(SquareR, Smooth)
Else
AvgSqrR = SquareR;
Plot1(AvgSqrR, ÒAvgSqrRÓ);
Plot2(Trend, ÒTrendÓ);
Plot3(NoTrend, ÒNoTrendÓ);
IF MRO(AvgSqrR > Trend, RunLen, RunLen) <> -1 OR
MRO(AvgSqrR < NoTrend, RunLen, RunLen) <> -1 Then Begin
Plot4(AvgSqrR, ÒRunningÓ);
Alert = True;
End;
Style:
| Plot |
Name |
Type |
Color |
Weight |
| Plot1 |
AvgSqrR |
Line |
Cyan |
thinnest |
| Plot2 |
Trend |
Line |
red |
thinnest |
| Plot3 |
NoTrend |
Line |
red |
thinnest |
| Plot4 |
Running |
Point |
white |
medium |
| Scaling: |
Screen |
|
|
|
| Properties: |
Enable Alert |
|
|
|
The PaintBar studies paint the bars that qualify as trending or nontrending. One PaintBar study paints the bars that are indicated to be trending, the other paints the nontrending. I found that all bars could be classified as trending or nontrending by setting the Trend and NoTrend inputs in both PaintBar studies to the same value. Although the default values do not reflect this type of setup, they can easily be modified to reflect such settings, since they are input values. The PaintBar studies can be developed using the EasyLanguage code that follows. The suggested format settings are given after the EasyLanguage.
Type: PaintBar
Name: R-Squared Trend
Inputs: Length(30), Smooth(1), Trend(.7);
Vars: R(0), SquareR(0), AvgSqrR(0);
R = coeffR(Length);
SquareR = Square(R);
IF Smooth > 0 Then
AvgSqrR = Average(SquareR, Smooth)
Else
AvgSqrR = SquareR;
IF AvgSqrR > Trend Then Begin
Plot1(High, ÒPaint_TrendÓ);
Plot2(Low, ÒPaint_TrendÓ);
Alert = True;
End;
Style:
Color: Cyan
Properties:
Enable Alert
Type: PaintBar
Name: R-Squared NoTrend
Inputs: Length(30), Smooth(1), NoTrend(.3);
Vars: R(0), SquareR(0), AvgSqrR(0);
R = coeffR(Length);
SquareR = Square(R);
IF Smooth > 0 Then
AvgSqrR = Average(SquareR, Smooth)
Else
AvgSqrR = SquareR;
IF AvgSqrR < NoTrend Then Begin
Plot1(High, ÒPaintNoTrendÓ);
Plot2(Low, ÒPaintNoTrendÓ);
Alert = True;
End;
Style:
Color: Red
Properties:
Enable Alert
This code is available at Omega Research's Web site. The name of the file is "RSqrd.ELA." Please note that all Traders' Tips analysis techniques posted at Omega Research's Web site can be utilized by both TradeStation and SuperCharts. Whenever possible, the posted analysis techniques will include both Quick Editor and Power Editor formats.
Gaston Sanchez, Omega Research
800 422-8587, 305 270-1095
Internet: https://www.omegaresearch.com
GO BACK
Here is a WAVE WI$E program implementation of Jack Karczewski's "Identifying market trends."
WAVE WI$E Spreadsheet Formulas
A: DATE @TC2000(c:\tc2000\data,
SP-500,STANDARD&POORS500,DB)
B: HIGH
C: LOW
D: CLOSE
E: OPEN
F: VOL
G: Lreg @LREG(CLOSE,30,0) 'linearregression
H: Slope @LREG(CLOSE,30,1) 'linearregressionslope
I: Rsq @LREG(CLOSE,30,2) 'Rsquared
J: count @IF(RSQ>.7,@CUM(1)) 'countconsecutivedaysRsq>.7
K:
==========End Formulas
--Peter Di Girolamo, Jerome Technology
908 369-7503, E-mail: jtiware@aol.com
Internet: https://members.aol.com/jtiware
GO BACK
In "Identifying market trends," Jack Karczewski describes several standard statistical trend indicators and their relationship to spotting trends. Specifically, he describes r-squared, a predicted value (based on a least-squares fit) and the error associated with this prediction. This can be implemented in Window On WallStreet Day Trader and Window On WallStreet Professional Investor.
In Window On WallStreet, begin with any open chart on your screen. Click the Indicator button on the toolbar, click New and enter the following formulas:
Name: R-Squared
Description: Jack Karczewski's R-Squared from Technical Analysis of Stocks and Commodities.
Variables:
Name: pers
Prompt Text: Enter Number of Period for Correlation
Default Value: 10
Formula(s):
pwr(correl(c,RecIndex(),pers,0),2)
Predicted Value based on a least squares fit
Name: Linear Predicted Value
Description: Jack Karczewski's linear predicted value from Technical
Analysis of Stocks and Commodities.
Variables:
Name: pers
Prompt Text: Enter number of periods for trend
Default Value: 10
Formula(s):
ref(c,-1) +ref(slope(c,pers),-1)
Percent Error of the linear predicted value
Name: Percent Error of LPV
Description: Percent error of Jack Karczewski's linear predicted value from
Technical Analysis of Stocks and Commodities.
Variables:
Name: pers
Prompt Text: Enter number of periods for trend
Default Value: 10
Formula(s):
100*(c- ref(c,-1) + ref(slope(c,pers),-1))/( ref(c,-1)
+ref(slope(c,pers),-1) )
--Andrew E. Laska, Window On WallStreet
888 732-5969, 972 783-6792
Internet: https://www.wallstreet.net
TRADESTATION / SUPERCHARTS 
TRADESTATION / SUPERCHARTS
