Corrected Moving Average

[helmutw]

The following Code of a moving average is published in a bulletin board by Andreas Uhl.

I can´t "read" a Metastock Code. Is it possible to code this in eSignal efs?


Thank You!

Helmut





{Corrected Average (CA), A.Uhl, Oct. 25, 2005}

vars:
CA(0),SA(0),n(0),
v1(0),v2(0),K(0);

inputs:
Price(Close),length(35);

if CurrentBar=1 then
CA=Price
else
begin

if CurrentBar<length then
n=CurrentBar
else
n=length;

SA=Average(Price,n);
v1=Square(StdDev(Price,n));
v2=Square(CA[1]-SA);

if v2<v1 then
K=0
else
K=1-v1/v2;
CA=CA[1]+K*(SA-CA[1]);
end;

Plot1(CA,"Corrected Average",Red,White,3);
Plot2(SA,"Simple Average",Blue,White,1);

[ACM]

Helmut
As I think the study could be of interest to this community I went ahead and coded the indicator. You can find a basic version of the study here
Alex

[helmutw]

Alex,

great, thank you for this!

Helmut

[helmutw]

A. Uhl published another MA- concept .

Maybe this study could also be of interest to this community.

Mr. Uhl was so kind to give me C++ Code.


// local function AEPMA
// copyright (c) A. Uhl 2005
// zero lag if input linear, that is line with constant slope

void AEPMAfun (const MSXDataInfoRec *a_psSrc, MSXDataInfoRec *a_psRslt)
{
int l_iIndex = a_psSrc->iFirstValid;
int l_iMaxIndex = a_psSrc->iLastValid;

float l_fSLOPE = 0 ; // slope of regression function
float l_fTAU = 1; // new time value
float l_fTAU0; // last time value
float l_fDTAU; // variable time step
float l_fDT = 1; // constant data time step (sample rate)
float l_fR0; // prediction one time step ahead
float l_fRX; // data regression
float l_fWAVG; // weighted average
float l_fYOLD; // last filter value
float l_fYNEW; // new filter value
float l_fQREST; // sum of squares of residuals
float l_fQTOT; // sum of squares of residuals
float l_fG; // gain factor
float l_fSX0; // sum of x0
float l_fSX0S; // sum of x0 squared
float l_fK = 0; // measure of determinism

int l_iN; // period (variable)
int X; // x-axis in regression


a_psRslt->pfValue[l_iIndex] = a_psSrc->pfValue[l_iIndex];

l_iIndex += 1;

while (l_iIndex <= l_iMaxIndex) {

l_fTAU0 = l_fTAU + l_fDT;
l_fYOLD = a_psRslt->pfValue[l_iIndex-1];
l_fR0 = l_fYOLD + l_fK*l_fSLOPE*l_fTAU0;
l_iN = int(floor(l_fTAU0));

l_fSX0 = 0;
l_fSX0S = 0;
for(X=0; X<=l_iN; X++) {
l_fSX0 += (l_fTAU0 - X);
l_fSX0S += pow(l_fTAU0 - X,2);
}

l_fWAVG=0;
for(X=0; X<=l_iN; X++) {
if(X==0) l_fRX = l_fR0;
else l_fRX = a_psSrc->pfValue[l_iIndex-X+1];
l_fWAVG += (l_fTAU0 - X) * l_fRX;
}

l_fWAVG = l_fWAVG / l_fSX0;
l_fSLOPE = (l_fWAVG - l_fYOLD) * l_fSX0 / l_fSX0S;

l_fQREST = 0;
l_fQTOT = 0;
for(X=0; X<=l_iN; X++) {
if(X==0) l_fRX = l_fR0;
else l_fRX = a_psSrc->pfValue[l_iIndex-X+1];
l_fQREST += pow(l_fRX - (l_fSLOPE*(l_fTAU0-X)+l_fYOLD),2);
l_fQTOT += pow(l_fRX - l_fYOLD,2);
}

l_fK = 1 - l_fQREST/l_fQTOT;
l_fDTAU = l_fK*(l_fTAU0-1);
l_fG = l_fDTAU*l_fSX0/l_fSX0S;
l_fTAU = l_fTAU0 - l_fDTAU;

// AEPMA
l_fYNEW = l_fYOLD + l_fG*(l_fWAVG-l_fYOLD);
a_psRslt->pfValue[l_iIndex] = l_fYNEW;

l_iIndex++;
}