mirror of
https://github.com/mihakralj/QuanTAlib.git
synced 2026-08-17 01:58:06 +00:00
docs: update category index files and fix indicator implementations (#58)
This commit is contained in:
Miha Kralj
+21
-60
@@ -1,67 +1,28 @@
|
||||
# Filters
|
||||
|
||||
> "All moving averages are low-pass filters. The question is which trade-offs you accept." John Ehlers
|
||||
> "All moving averages are low-pass filters. The question is which trade-offs you accept." John Ehlers
|
||||
|
||||
Signal processing filters adapted for financial time series. These are not indicators in the traditional sense: they are building blocks. Low-pass removes noise. High-pass isolates cycles. Band-pass extracts specific frequencies. Each filter type trades off smoothness, lag, and overshoot differently.
|
||||
|
||||
## Indicator Status
|
||||
## Indicators
|
||||
|
||||
| Indicator | Full Name | Status | Description |
|
||||
| :--- | :--- | :---: | :--- |
|
||||
| [Bessel](lib/filters/bessel/Bessel.md) | Bessel Filter | | Maximally flat group delay. Best phase response. Minimal overshoot. |
|
||||
| [Bilateral](lib/filters/bilateral/Bilateral.md) | Bilateral Filter | | Edge-preserving smoothing. Adapts to local gradients. |
|
||||
| [BPF](lib/filters/bpf/Bpf.md) | BandPass Filter | | 2nd-order IIR. Cascade of HP + LP. Extracts specific frequency band. |
|
||||
| [Butter](lib/filters/butter/Butter.md) | Butterworth Filter | | Maximally flat frequency response. Classic IIR filter. |
|
||||
| Cheby1 | Chebyshev Type I | =Ë | Steeper roll-off with passband ripple. Sharper cutoff than Butterworth. |
|
||||
| Cheby2 | Chebyshev Type II | =Ë | Equiripple stopband, monotonic passband. Better stopband rejection. |
|
||||
| [Elliptic](lib/filters/elliptic/Elliptic.md) | Elliptic Filter | | Equiripple both bands. Sharpest transition for given order. |
|
||||
| [Gauss](lib/filters/gauss/Gauss.md) | Gaussian Filter | | Bell-curve weighted smoothing. No overshoot. |
|
||||
| [Hann](lib/filters/hann/Hann.md) | Hann Filter | | Hann window smoothing. Good spectral leakage control. |
|
||||
| [Hp](lib/filters/hp/Hp.md) | Hodrick-Prescott | | Causal trend/cycle decomposition. Regularization parameter » controls smoothness. |
|
||||
| [Hpf](lib/filters/hpf/Hpf.md) | High Pass Filter | | Attenuates below cutoff. Isolates fast components. |
|
||||
| [Kalman](lib/filters/kalman/Kalman.md) | Kalman Filter | | Recursive state estimation. Optimal under Gaussian assumptions. |
|
||||
| [Loess](lib/filters/loess/Loess.md) | LOESS Smoothing | | Local polynomial regression. Robust to outliers. |
|
||||
| [Notch](lib/filters/notch/Notch.md) | Notch Filter | | Band-stop. Removes specific frequency (e.g., 60 Hz noise). |
|
||||
| [SGF](lib/filters/sgf/Sgf.md) | Savitzky-Golay | | Polynomial smoothing. Preserves higher moments (derivatives). |
|
||||
| [SSF](lib/filters/ssf/Ssf.md) | Super Smoother | | Ehlers. 2-pole Butterworth variant. Standard cycle pre-filter. |
|
||||
| [USF](lib/filters/usf/Usf.md) | Ultra Smoother | | Ehlers. 3-pole variant. More smoothing than SSF. |
|
||||
| Wiener | Wiener Filter | =Ë | Optimal linear filter. Minimizes MSE given signal/noise spectra. |
|
||||
|
||||
**Status Key:** Implemented | =Ë Planned
|
||||
|
||||
## Selection Guide
|
||||
|
||||
| Use Case | Recommended | Why |
|
||||
| Indicator | Full Name | Description |
|
||||
| :--- | :--- | :--- |
|
||||
| General smoothing | Butter, SSF | Good balance of smoothing and lag. |
|
||||
| Minimal overshoot | Bessel, Gauss | Bessel: best phase. Gauss: no overshoot by design. |
|
||||
| Sharp cutoff | Elliptic, Cheby1 | Elliptic: sharpest. Cheby1: simpler. |
|
||||
| Cycle extraction | BPF, Hp | BPF for specific band. Hp for trend/cycle split. |
|
||||
| Noise spike removal | Notch | Surgical removal of specific frequency. |
|
||||
| Outlier robustness | Bilateral, Loess | Adapt to local structure. Ignore outliers. |
|
||||
| Derivative preservation | SGF | Polynomial fit preserves shape. |
|
||||
| Adaptive estimation | Kalman | Updates estimate as new data arrives. Optimal under model. |
|
||||
|
||||
## Filter Characteristics
|
||||
|
||||
| Filter | Type | Order | Overshoot | Lag | Sharpness |
|
||||
| :--- | :--- | :---: | :---: | :---: | :---: |
|
||||
| Butter | IIR LP | 2 | Low | Medium | Medium |
|
||||
| Bessel | IIR LP | 2 | Minimal | Higher | Low |
|
||||
| Cheby1 | IIR LP | 2 | Higher | Lower | High |
|
||||
| Elliptic | IIR LP | 2 | Higher | Lowest | Highest |
|
||||
| SSF | IIR LP | 2 | Low | Low | Medium |
|
||||
| USF | IIR LP | 3 | Lower | Medium | Medium |
|
||||
| Gauss | FIR LP | N | None | Higher | Low |
|
||||
| SGF | FIR LP | N | Low | Medium | Low |
|
||||
|
||||
Higher order = more smoothing but more lag. IIR filters have minimal coefficients but can overshoot. FIR filters are always stable with linear phase but need more coefficients.
|
||||
|
||||
## Filter Design Principles
|
||||
|
||||
| Principle | Trade-off | QuanTAlib Approach |
|
||||
| :--- | :--- | :--- |
|
||||
| Smoothness vs lag | More smoothing = more lag | Parameterized period/cutoff |
|
||||
| Sharpness vs ripple | Sharper cutoff = more ripple | Choose filter type for application |
|
||||
| Stability | IIR can be unstable | All implementations verified stable |
|
||||
| Causality | Real-time requires causal filters | All filters are causal (no lookahead) |
|
||||
| [BESSEL](lib/filters/bessel/Bessel.md) | Bessel Filter | Maximally flat group delay. Best phase response. Minimal overshoot. |
|
||||
| [BILATERAL](lib/filters/bilateral/Bilateral.md) | Bilateral Filter | Edge-preserving smoothing. Adapts to local gradients. |
|
||||
| [BPF](lib/filters/bpf/Bpf.md) | BandPass Filter | 2nd-order IIR. Cascade of HP + LP. Extracts specific frequency band. |
|
||||
| [BUTTER](lib/filters/butter/Butter.md) | Butterworth Filter | Maximally flat frequency response. Classic IIR filter. |
|
||||
| [CHEBY1](lib/filters/cheby1/Cheby1.md) | Chebyshev Type I | Steeper roll-off with passband ripple. Sharper cutoff than Butterworth. |
|
||||
| [CHEBY2](lib/filters/cheby2/Cheby2.md) | Chebyshev Type II | Equiripple stopband, monotonic passband. Better stopband rejection. |
|
||||
| [ELLIPTIC](lib/filters/elliptic/Elliptic.md) | Elliptic Filter | Equiripple both bands. Sharpest transition for given order. |
|
||||
| [GAUSS](lib/filters/gauss/Gauss.md) | Gaussian Filter | Bell-curve weighted smoothing. No overshoot. |
|
||||
| [HANN](lib/filters/hann/Hann.md) | Hann Filter | Hann window smoothing. Good spectral leakage control. |
|
||||
| [HP](lib/filters/hp/Hp.md) | Hodrick-Prescott | Causal trend/cycle decomposition. Regularization parameter λ controls smoothness. |
|
||||
| [HPF](lib/filters/hpf/Hpf.md) | High Pass Filter | Attenuates below cutoff. Isolates fast components. |
|
||||
| [KALMAN](lib/filters/kalman/Kalman.md) | Kalman Filter | Recursive state estimation. Optimal under Gaussian assumptions. |
|
||||
| [LOESS](lib/filters/loess/Loess.md) | LOESS Smoothing | Local polynomial regression. Robust to outliers. |
|
||||
| [NOTCH](lib/filters/notch/Notch.md) | Notch Filter | Band-stop. Removes specific frequency (e.g., 60 Hz noise). |
|
||||
| [SGF](lib/filters/sgf/Sgf.md) | Savitzky-Golay | Polynomial smoothing. Preserves higher moments (derivatives). |
|
||||
| [SSF](lib/filters/ssf/Ssf.md) | Super Smoother | Ehlers. 2-pole Butterworth variant. Standard cycle pre-filter. |
|
||||
| [USF](lib/filters/usf/Usf.md) | Ultra Smoother | Ehlers. 3-pole variant. More smoothing than SSF. |
|
||||
| [WIENER](lib/filters/wiener/Wiener.md) | Wiener Filter | Optimal linear filter. Minimizes MSE given signal/noise spectra. |
|
||||
|
||||
@@ -113,10 +113,7 @@ public class SgfValidationTests : IDisposable
|
||||
if (wSum <= double.Epsilon)
|
||||
{
|
||||
int availablePoints = Math.Min(i + 1, adjPeriod);
|
||||
if (availablePoints < adjPeriod)
|
||||
result[i] = source[i];
|
||||
else
|
||||
result[i] = double.NaN;
|
||||
result[i] = availablePoints < adjPeriod ? source[i] : double.NaN;
|
||||
}
|
||||
}
|
||||
|
||||
@@ -202,4 +199,4 @@ public class SgfValidationTests : IDisposable
|
||||
}
|
||||
_output.WriteLine("Span mode successfully validated against reference implementation");
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
@@ -248,11 +248,8 @@ public sealed class Sgf : AbstractBase
|
||||
}
|
||||
else
|
||||
{
|
||||
// If wSum is zero/negative/small
|
||||
if (count < adjPeriod)
|
||||
output[i] = source[i]; // Pass through for partial window
|
||||
else
|
||||
output[i] = double.NaN;
|
||||
// If wSum is zero/negative/small - pass through for partial window, else NaN
|
||||
output[i] = count < adjPeriod ? source[i] : double.NaN;
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
Reference in New Issue
Block a user