July 3rd, 2010
Size: 6.55 KB
Downloaded: 174 times
521
fluxsmith
Linear Regression Intercept. Returns slightly different values than the standard LinRegIntercept provided, which I believe to be incorrect (at least through NT7 Beta 18). Also unlike the provided version does not require iteration on intrabar updates. Version 1.0.
July 3rd, 2010
Size: 4.59 KB
Downloaded: 162 times
520
fluxsmith
R-Squared, Coefficient of Determination, the square of the Correlation Coefficient. This should return the same values as the standard supplied RSquared. Unfortunately both versions always iterate, so should probably be used with CalculateOnBarClose == true.
I believe my version to be slightly more CPU efficient, as it does not require a square root calculation used in the distribution version.
Version 1.0.
July 3rd, 2010
Size: 6.78 KB
Downloaded: 164 times
525
fluxsmith
Linear Regression Slope. Should return the same values as the standard LinRegSlope. However, the standard version iterates on every update, making it especially inefficient for intrabar (tick) data. This version does not iterate on intrabar updates. Version 1.0.
I just stood on his shoulders and converted it to my style. This version uses much less memory, is simpler for me to follow, and has a startup phase where it has a pretty good estimate instead of no value set.
It can be constructed from my generic MA adaptor (jhlMA) with only a period specified, providing a default phase value of 0.
February 12th, 2011
Size: 15.66 KB
Downloaded: 335 times
827
fluxsmith
Maximum value seen in period. Should return the same values as the standard MAX, which as of b18 has been fixed to reduce iteration. With that fixed the only advantage this has over the standard is memory reduction when used as a component and not displayed. (Both versions are now efficient with CalculateOnBarClose == false.)
Version 2 - Corrected calculation errors when COBC == false && Input != High
July 4th, 2010
Size: 2.31 KB
Downloaded: 278 times
527
fluxsmith
I came across "McGinley Dynamic" in a thread on this forum. On further investigation I found two entirely different formulas referred to by that name. This indicator implements a variation based on the following formula:
v[0] = v[1] + (Close[0] - v[1]) / (.618 * n * (Close[0] / v[1]) ^ 4)
Now that I can see the result, it doesn't appear to me to be any more useful as an average than an EMA of the same period. However I think it may make a good trigger line vs an EMA of the same period.