反正态累积分布函数转换

Question

we used the Extreme Numerics library from Extreme Optimization ( http://www.extremeoptimization.com ) for our calculation kernel.

// portfolio risk is inverse normal cumulative distribution function
double risk = NormalDistribution.InverseDistributionFunction(
  _riskProbabilityLevel, 
  mean, 
  stdev);

The problem is now we are moving from C# to Java and I don't really know all that much about Java but have been tasked with re-writing this particular function.

I have values to test against:

RiskProbabilityLevel = 0.02
Mean                 = 0.06618
Standard Dev         = 0.057196166520267355

Risk                 = 0.051286461995869864

but in looking thru the various functions in math3.distribution.NormalDistribution libraries I can't find what might be equivalent.

Any direction or help would be appreciated. thanks.

Answer 1

I think you can use this library. It's called apache commons.math3 . It's very popular. From the docs supplied by http://www.extremeoptimization.com I think you can use the following code:

import org.apache.commons.math3.distribution.NormalDistribution;

public class TestProbabilites {

    public static void main(String[] args) {
        double riskProbabilityLevel = 0.02D;
        double mean = 0.06618D;
        double standardDev = 0.057196166520267355D;
        double expectedRisk = 0.051286461995869864D;

        NormalDistribution distribution = new NormalDistribution(mean, standardDev);
        double outcomeRisk = distribution.inverseCumulativeProbability(riskProbabilityLevel);
    }
}

But you have to remember to add the apache.commons.math3 library added to your project using Maven or Gradle or some other dependencies manager.

Hopes this helps.