LinRegFit - Wealth Lab Forum

mikael

#1

2/19/2009 8:32 PM

For your viewing pleasure.

CODE:
Please log in to see this code.

Size:

Color:

For your viewing pleasure.

[CODE]
    static public List<double> LinRegFit( List<Point> points )
    {
        // Check Args
        if( points.Count < 3 )
           throw new Exception( "Too few points." );

// Init vars
        int i; // used for indexing
        double t; // a temporary hold variable

double sum_x = 0.0, sum_y = 0.0, st2 = 0.0;

for( i= 0 ; i < points.Count ; i++ )
        {
            sum_x += points[i].x;
            sum_y += points[i].y;
        }

double mean_x = sum_x / points.Count;
        double slope = 0.0;
        for( i= 0 ; i < points.Count ; i++ )
        {
            t = points[i].x - mean_x;
            st2 += t*t;
            slope += t * points[i].y;
        }

slope /= st2;
        double intercept = (sum_y - sum_x * slope) / points.Count;
        double sigma_intercept = 0.0, sigma_slope = 0.0; 
        if( st2 != 0.0 )
        {
            sigma_intercept = Math.Sqrt((1.0 + sum_x * sum_x / (points.Count * st2)) / points.Count);
            sigma_slope = Math.Sqrt(1.0 / st2);
        }

double chi_squared = 0.0;
        for( i= 0 ; i < points.Count ; i++ )
        {
            t = points[i].y - intercept - slope * points[i].x;
            chi_squared = chi_squared + t * t;
        }

// need x.Count() >= 3 here
        t = Math.Sqrt(chi_squared / (points.Count - 2));
        sigma_intercept = sigma_intercept * t;
        sigma_slope = sigma_slope * t;

// Results List
        List<double> result = new List<double>();
        result.Add(intercept);    result.Add(slope);
        result.Add(sigma_intercept); result.Add(sigma_slope);
        result.Add(chi_squared);

return result;
    }
[/CODE]

TrendCatcher

#2

2/19/2009 9:08 PM

Hi Mikael,
I don't see anything on the chart. Is there any error in the script?

DrKoch

#3

2/20/2009 1:22 AM

mikael,
you translated my WL4 study "LinRegFit" to WL5. Good.
Please give credits to the original author (or at least mention where you take your ideas from)

TrendCatcher,
LinRegFit is just a helper function which can do a linear regression on a collection of points, very helpful sometimes.

TrendCatcher

#4

2/20/2009 7:41 AM

Hi DrKoch,
I don't see any regression lines on the chart. Could you post the entire script if the above is not complete? Thank you.

mikael

#5

2/20/2009 11:43 AM

Sorry DrKoch. I wasn't claiming originality just providing the code. So, yes refined and translated from Dr Koch's original Object Pascal code into C#. And originally I think from "Numerical Recipes in C". That do?

BTW you're going to need this class for the above:

CODE:
Please log in to see this code.

jimcrist

#6

3/5/2009 9:08 PM

I want to use this routine to get the slope and therefore the trend. I thought I had this working in my program, but the slope is always negative?!? Or always positive if X goes from 10 down to 1. I'm using the daily S&P500 as the index. I've checked sigma slope and slope (what's the difference?). The data works back from the date shown.

11/15/07 downtrend
i=23323 X=1 MClose[i]=1451.15
i=23297 X=2 MClose[i]=1470.58
i=23271 X=3 MClose[i]=1481.05
i=23245 X=4 MClose[i]=1439.1801
i=23219 X=5 MClose[i]=1453.7
i=23193 X=6 MClose[i]=1474.77
i=23167 X=7 MClose[i]=1475.62
i=23141 X=8 MClose[i]=1520.27
i=23115 X=9 MClose[i]=1502.17
i=23089 X=10 MClose[i]=1509.65
Slope=-0.0980010101009887

11/13/2006 uptrend
i=16764 X=1 MClose[i]=1384.42
i=16738 X=2 MClose[i]=1380.9
i=16712 X=3 MClose[i]=1378.33
i=16686 X=4 MClose[i]=1385.72
i=16660 X=5 MClose[i]=1382.84
i=16634 X=6 MClose[i]=1379.78
i=16608 X=7 MClose[i]=1367.34
i=16582 X=8 MClose[i]=1367.8101
i=16556 X=9 MClose[i]=1377.9399
i=16530 X=10 MClose[i]=1377.9301
Slope=-3.40190969696967

11/11/04 uptrend
i=3854 X=1 MClose[i]=1173.48
i=3828 X=2 MClose[i]=1162.91
i=3802 X=3 MClose[i]=1164.08
i=3776 X=4 MClose[i]=1164.89
i=3750 X=5 MClose[i]=1161.67
i=3724 X=6 MClose[i]=1143.2
i=3698 X=7 MClose[i]=1130.54
i=3672 X=8 MClose[i]=1130.51
i=3646 X=9 MClose[i]=1130.2
i=3620 X=10 MClose[i]=1127.4399
Slope=-5.65273272727273

CODE:
Please log in to see this code.

Note: I altered LinRegFit to just return the slope.

Size:

Color:

I want to use this routine to get the slope and therefore the trend. I thought I had this working in my program, but the slope is always negative?!? Or always positive if X goes from 10 down to 1. I'm using the daily S&P500 as the index. I've checked sigma slope and slope (what's the difference?). The data works back from the date shown.

11/15/07 downtrend
i=23323 X=1 MClose[i]=1451.15
i=23297 X=2 MClose[i]=1470.58
i=23271 X=3 MClose[i]=1481.05
i=23245 X=4 MClose[i]=1439.1801
i=23219 X=5 MClose[i]=1453.7
i=23193 X=6 MClose[i]=1474.77
i=23167 X=7 MClose[i]=1475.62
i=23141 X=8 MClose[i]=1520.27
i=23115 X=9 MClose[i]=1502.17
i=23089 X=10 MClose[i]=1509.65
Slope=-0.0980010101009887

11/13/2006 uptrend
i=16764 X=1 MClose[i]=1384.42
i=16738 X=2 MClose[i]=1380.9
i=16712 X=3 MClose[i]=1378.33
i=16686 X=4 MClose[i]=1385.72
i=16660 X=5 MClose[i]=1382.84
i=16634 X=6 MClose[i]=1379.78
i=16608 X=7 MClose[i]=1367.34
i=16582 X=8 MClose[i]=1367.8101
i=16556 X=9 MClose[i]=1377.9399
i=16530 X=10 MClose[i]=1377.9301
Slope=-3.40190969696967

11/11/04 uptrend
i=3854 X=1 MClose[i]=1173.48
i=3828 X=2 MClose[i]=1162.91
i=3802 X=3 MClose[i]=1164.08
i=3776 X=4 MClose[i]=1164.89
i=3750 X=5 MClose[i]=1161.67
i=3724 X=6 MClose[i]=1143.2
i=3698 X=7 MClose[i]=1130.54
i=3672 X=8 MClose[i]=1130.51
i=3646 X=9 MClose[i]=1130.2
i=3620 X=10 MClose[i]=1127.4399
Slope=-5.65273272727273
[CODE]			int X = 1;
			for (int i = Bar-BarsInDay; i >= Bar - (MktIndxLngth*BarsInDay); i -= BarsInDay)
			{
				Point myPoint = new Point(X, MClose[i]);
				myList.Add(myPoint);
				if (Debug) PrintDebug("i=" + i + " X=" + X + " MClose[i]=" + MClose[i]);
				X = X + 1;
			}
			Slope = LinRegFit(myList);
			if (Debug) PrintDebug("Slope=" + Slope);
[/CODE]
Note: I altered LinRegFit to just return the slope.

jimcrist

#7

3/6/2009 9:33 AM

Nevermind, I guess it's working...

sgubba

#8

3/6/2009 12:42 PM

Hi Eugene,
I don't see anything happening on the chart. Can you re-post the script with corrected cod?

Eugene

#9

3/6/2009 1:03 PM

Why not ask its author? Neither it's my code nor I have the interest to follow it.

sgubba

#10

3/6/2009 1:11 PM

:)

jimcrist

#11

3/6/2009 8:23 PM

squbba: if you want to see a regression line on a chart I think you want this thread:

http://wealth-lab.com/Community/Forum/Posts.aspx?id=gomhZfqEsRslInG2b1zSNe5rZSd5IlLXT0KDK0odA9lwY4TOu4PSXYE9KKyGtbkxUQ/6ouW5dPYXKNtU26gPNaizZrzlGHgyl9V0YdCTLS4=

sgubba

#12

3/6/2009 9:15 PM

Jimcrist,
Thank you very much.

Eugene

#13

3/7/2009 1:03 AM

In addition, you may want to take a look at TSF (Time Series Forecast) in Community.Indicators.

sgubba

#14

3/7/2009 8:58 PM

Thank you Eugene.

zanek

#15

6/24/2009 12:24 PM

Eugene,

I tried to add : "using Community.Indicators;"

to my code but it says the type or namespace couldnt be found. What do I need to do so that that namespace can be found ?