Class VarianceRTIndicator

java.lang.Object

com.wualabs.qtsurfer.engine.indicators.core.AbstractRTIndicator

com.wualabs.qtsurfer.engine.indicators.core.AbstractIncrementalRTIndicator

com.wualabs.qtsurfer.engine.indicators.core.AbstractDecorableIncrementalRTIndicator

com.wualabs.qtsurfer.engine.indicators.statistics.VarianceRTIndicator

All Implemented Interfaces:

ClockAware, Resettable, RTIndicator

public class VarianceRTIndicator
extends AbstractDecorableIncrementalRTIndicator

Variance RTIndicator

In periodic mode, by default (VariancePrecision.EXACT), computes the variance of the window from the window itself on every tick: the mean first, then the squared deviations from it, oldest value first. That is O(periods) per tick, exact to the last digits, and there is nothing to drift. It replaces a running sum of squares with Var(X) = E(X²) - E(X)², which cancels the deviation away when the mean is large next to it (a quiet window of prices in the tens of thousands) and keeps the rounding of every add and eviction for good.

withPrecision(FAST) trades the last digits for O(1) per tick: running sums of the deviations from a reference value near the window's mean, with a bound on the rounding they gather and an exact pass over the window whenever that bound could matter. It keeps the rounding of the variance below about one part in a thousand million (the bound is a worst case; the tests hold it to ten times that), wherever the mean is.

Constructor Summary

Constructors

Constructor	Description
VarianceRTIndicator()
VarianceRTIndicator(int periods)
VarianceRTIndicator(RTIndicator indicator)
VarianceRTIndicator(RTIndicator indicator, int periods)

Method Summary

Modifier and Type	Method	Description
int	getFastRecalculations()	How many times FAST worked the window out again from the window itself (for tests and diagnostics).
VariancePrecision	getPrecision()
SmaRTIndicator	getSma()
protected double	nextValue(double prevValue, double newValue)	Non-periodic (2-period) variance: running average of squared deviations.
void	reset()
protected double	tickValue(int tick, boolean init, boolean endPeriod, double newValue)	Periodic variance computation from the window: mean = Σx/n, then Σ(x-mean)²/n (population) or Σ(x-mean)²/(n-1) with Bessel's correction (sample), the sums added in the order the values arrived.
VarianceRTIndicator	withAverageSample(boolean averageSample)
VarianceRTIndicator	withPrecision(VariancePrecision precision)	Chooses how the variance is worked out.

Methods inherited from class AbstractDecorableIncrementalRTIndicator

getIndicator, isReady, setClockSupplier, update

Methods inherited from class AbstractIncrementalRTIndicator

getLastUpdateValue, getPeriodCycles, getPeriods, getPrevValue, initValue, isPeriodic, prevNextValue, setPeriods

Methods inherited from class AbstractRTIndicator

equals, getDisplayHint, getMeta, getValue, hashCode, hide, isPoison, setValue, toString, toString, withDisplayHint, withMeta

Methods inherited from class Object

clone, finalize, getClass, notify, notifyAll, wait, wait, wait

Methods inherited from interface RTIndicator

getId, isHidden, ro, update, update

Constructor Details

VarianceRTIndicator

public VarianceRTIndicator()

VarianceRTIndicator

public VarianceRTIndicator(int periods)

VarianceRTIndicator

public VarianceRTIndicator(RTIndicator indicator)

VarianceRTIndicator

public VarianceRTIndicator(RTIndicator indicator,
 int periods)

Method Details

reset

public void reset()

Specified by:

reset in interface Resettable

Overrides:

reset in class AbstractDecorableIncrementalRTIndicator

withAverageSample

public VarianceRTIndicator withAverageSample(boolean averageSample)

withPrecision

public VarianceRTIndicator withPrecision(VariancePrecision precision)

Chooses how the variance is worked out. VariancePrecision.EXACT (the default) works it out from the window on every tick; VariancePrecision.FAST keeps O(1) running sums with a guard. Safe to change at any point: FAST starts from the values the window holds at that moment.

tickValue

protected double tickValue(int tick,
 boolean init,
 boolean endPeriod,
 double newValue)

Periodic variance computation from the window: mean = Σx/n, then Σ(x-mean)²/n (population) or Σ(x-mean)²/(n-1) with Bessel's correction (sample), the sums added in the order the values arrived. Returns 0 until the first complete period.

Overrides:

tickValue in class AbstractIncrementalRTIndicator

Parameters:

tick - current tick within the period

init - true on the very first update

endPeriod - true when this tick completes a period

newValue - the new input value

Returns:

the computed variance, or 0 if insufficient data

nextValue

protected double nextValue(double prevValue,
 double newValue)

Non-periodic (2-period) variance: running average of squared deviations.

Overrides:

nextValue in class AbstractDecorableIncrementalRTIndicator

Parameters:

prevValue - the previous variance value

newValue - the new input value

Returns:

the smoothed variance

getSma

public SmaRTIndicator getSma()

getPrecision

public VariancePrecision getPrecision()

getFastRecalculations

public int getFastRecalculations()

How many times FAST worked the window out again from the window itself (for tests and diagnostics).