siliconhippy 37 posts msg #32235  Ignore siliconhippy 
6/2/2004 11:41:59 PM
Hey Folks,
Was trying to write an Rsquared (coefficient of determination)filter as below, but must have had syntax errors (0 matches.)
Rsq values vary between 1 and 0, 1 being a perfect straight line chart.
See legend after filter. Please help !
Rsq>0.90UpVol FILTER

set{avgvalue, sum(close,90)/90}
set{sqdiff, (close  avgvalue)*(close  avgvalue)}
set{SSE, sum(sqdiff, 90)}
set{SST, (90*standard deviation(90)*standard deviation(90))}
set{Rsq, (1 SSE/SST)}
show stocks where Rsq is above 0.90 and average volume is above 100000 and volume is above 125000 and volume has been increasing for 3 days

LEGEND:
Where Rsquared= 1  SSE/SST
SSE = Sum [(Y(i)avgY(i))**2] over all i=1,...,n, say n= last 90 days SST = Sum[Y(i)**2]  (1/n)* [SumY(i)}**2 avgY(i)= Sum Y(i)/n = SUM(close,90)/90 where n=90
Therefore, SST(n) = n * standard deviation(90)**2 n=90
SF has the above function, as STANDARD DEVIATION(period, avg period) and also SUM(IND, LENGTH)
Regards,
siliconhippy.

cegis 235 posts msg #32249  Ignore cegis 
6/3/2004 2:35:51 PM
siliconhippy,
If I'm not mistaken, you can only use one arithmetic operation per set{} command. Therefore, you'll need to break apart lines 2, 4 and 5 into multiple set{}'s. (Eg., for line 2 you'd need set{sd1,close  avgvalue} set{sqdiff, sqd1 * sqd1})
Oh, yeah. There's a rsquared function now, too!
HTH,
C
