Inverse Gaussian distribution

In probability theory, the inverse Gaussian distribution (also known as the Wald distribution) is a two-parameter family of continuous probability distributions with support on (0,∞).

Inverse Gaussian
Probability density function
Cumulative distribution function
Notation
Parameters
Support
PDF
CDF

where is the standard normal (standard Gaussian) distribution c.d.f.
Mean


Mode
Variance


Skewness
Excess kurtosis
MGF
CF

Its probability density function is given by

for x > 0, where is the mean and is the shape parameter.[1]

The inverse Gaussian distribution has several properties analogous to a Gaussian distribution. The name can be misleading: it is an "inverse" only in that, while the Gaussian describes a Brownian motion's level at a fixed time, the inverse Gaussian describes the distribution of the time a Brownian motion with positive drift takes to reach a fixed positive level.

Its cumulant generating function (logarithm of the characteristic function)[contradictory] is the inverse of the cumulant generating function of a Gaussian random variable.

To indicate that a random variable X is inverse Gaussian-distributed with mean μ and shape parameter λ we write .

Properties

edit

Single parameter form

edit

The probability density function (pdf) of the inverse Gaussian distribution has a single parameter form given by

 

In this form, the mean and variance of the distribution are equal,  

Also, the cumulative distribution function (cdf) of the single parameter inverse Gaussian distribution is related to the standard normal distribution by

 

where  ,   and the   is the cdf of standard normal distribution. The variables   and   are related to each other by the identity  

In the single parameter form, the MGF simplifies to

 

An inverse Gaussian distribution in double parameter form   can be transformed into a single parameter form   by appropriate scaling   where  

The standard form of inverse Gaussian distribution is

 

Summation

edit

If Xi has an   distribution for i = 1, 2, ..., n and all Xi are independent, then

 

Note that

 

is constant for all i. This is a necessary condition for the summation. Otherwise S would not be Inverse Gaussian distributed.

Scaling

edit

For any t > 0 it holds that

 

Exponential family

edit

The inverse Gaussian distribution is a two-parameter exponential family with natural parametersλ/(2μ2) and −λ/2, and natural statistics X and 1/X.

For   fixed, it is also a single-parameter natural exponential family distribution[2] where the base distribution has density

 

Indeed, with  ,

 

is a density over the reals. Evaluating the integral, we get

 

Substituting   makes the above expression equal to  .

Relationship with Brownian motion

edit
 
Example of stopped random walks with  . The upper figure shows the histogram of waiting times, along with the prediction according to inverse gaussian distribution. The lower figure shows the trajectories.

Let the stochastic process Xt be given by

 
 

where Wt is a standard Brownian motion. That is, Xt is a Brownian motion with drift  .

Then the first passage time for a fixed level   by Xt is distributed according to an inverse-Gaussian:

 

i.e

 

(cf. Schrödinger[3] equation 19, Smoluchowski[4], equation 8, and Folks[5], equation 1).

Derivation of the first passage time distribution

Suppose that we have a Brownian motion   with drift   defined by:

 

And suppose that we wish to find the probability density function for the time when the process first hits some barrier   - known as the first passage time. The Fokker-Planck equation describing the evolution of the probability distribution   is:

 

where   is the Dirac delta function. This is a boundary value problem (BVP) with a single absorbing boundary condition  , which may be solved using the method of images. Based on the initial condition, the fundamental solution to the Fokker-Planck equation, denoted by  , is:

 

Define a point  , such that  . This will allow the original and mirror solutions to cancel out exactly at the barrier at each instant in time. This implies that the initial condition should be augmented to become:

 

where   is a constant. Due to the linearity of the BVP, the solution to the Fokker-Planck equation with this initial condition is:

 

Now we must determine the value of  . The fully absorbing boundary condition implies that:

 

At  , we have that  . Substituting this back into the above equation, we find that:

 

Therefore, the full solution to the BVP is:

 

Now that we have the full probability density function, we are ready to find the first passage time distribution  . The simplest route is to first compute the survival function  , which is defined as:

 

where   is the cumulative distribution function of the standard normal distribution. The survival function gives us the probability that the Brownian motion process has not crossed the barrier   at some time  . Finally, the first passage time distribution   is obtained from the identity:

 

Assuming that  , the first passage time follows an inverse Gaussian distribution:

 

When drift is zero

edit

A common special case of the above arises when the Brownian motion has no drift. In that case, parameter μ tends to infinity, and the first passage time for fixed level α has probability density function

 

(see also Bachelier[6]: 74 [7]: 39 ). This is a Lévy distribution with parameters   and  .

Maximum likelihood

edit

The model where

 

with all wi known, (μλ) unknown and all Xi independent has the following likelihood function

 

Solving the likelihood equation yields the following maximum likelihood estimates

 

  and   are independent and

 

Sampling from an inverse-Gaussian distribution

edit

The following algorithm may be used.[8]

Generate a random variate from a normal distribution with mean 0 and standard deviation equal 1

 

Square the value

 

and use the relation

 

Generate another random variate, this time sampled from a uniform distribution between 0 and 1

 

If   then return   else return  

Sample code in Java:

public double inverseGaussian(double mu, double lambda) {
    Random rand = new Random();
    double v = rand.nextGaussian();  // Sample from a normal distribution with a mean of 0 and 1 standard deviation
    double y = v * v;
    double x = mu + (mu * mu * y) / (2 * lambda) - (mu / (2 * lambda)) * Math.sqrt(4 * mu * lambda * y + mu * mu * y * y);
    double test = rand.nextDouble();  // Sample from a uniform distribution between 0 and 1
    if (test <= (mu) / (mu + x))
        return x;
    else
        return (mu * mu) / x;
}
 
Wald distribution using Python with aid of matplotlib and NumPy

And to plot Wald distribution in Python using matplotlib and NumPy:

import matplotlib.pyplot as plt
import numpy as np

h = plt.hist(np.random.wald(3, 2, 100000), bins=200, density=True)

plt.show()
edit
  • If  , then   for any number  [1]
  • If   then  
  • If   for   then  
  • If   then  
  • If  , then  .[9]

The convolution of an inverse Gaussian distribution (a Wald distribution) and an exponential (an ex-Wald distribution) is used as a model for response times in psychology,[10] with visual search as one example.[11]

History

edit

This distribution appears to have been first derived in 1900 by Louis Bachelier[6][7] as the time a stock reaches a certain price for the first time. In 1915 it was used independently by Erwin Schrödinger[3] and Marian v. Smoluchowski[4] as the time to first passage of a Brownian motion. In the field of reproduction modeling it is known as the Hadwiger function, after Hugo Hadwiger who described it in 1940.[12] Abraham Wald re-derived this distribution in 1944[13] as the limiting form of a sample in a sequential probability ratio test. The name inverse Gaussian was proposed by Maurice Tweedie in 1945.[14] Tweedie investigated this distribution in 1956[15] and 1957[16][17] and established some of its statistical properties. The distribution was extensively reviewed by Folks and Chhikara in 1978.[5]

Rated Inverse Gaussian Distribution

edit

Assuming that the time intervals between occurrences of a random phenomenon follow an inverse Gaussian distribution, the probability distribution for the number of occurrences of this event within a specified time window is referred to as rated inverse Gaussian.[18] While, first and second moment of this distribution are calculated, the derivation of the moment generating function remains an open problem.

Numeric computation and software

edit

Despite the simple formula for the probability density function, numerical probability calculations for the inverse Gaussian distribution nevertheless require special care to achieve full machine accuracy in floating point arithmetic for all parameter values.[19] Functions for the inverse Gaussian distribution are provided for the R programming language by several packages including rmutil,[20][21] SuppDists,[22] STAR,[23] invGauss,[24] LaplacesDemon,[25] and statmod.[26]

See also

edit

References

edit
  1. ^ a b Chhikara, Raj S.; Folks, J. Leroy (1989), The Inverse Gaussian Distribution: Theory, Methodology and Applications, New York, NY, USA: Marcel Dekker, Inc, ISBN 0-8247-7997-5
  2. ^ Seshadri, V. (1999), The Inverse Gaussian Distribution, Springer-Verlag, ISBN 978-0-387-98618-0
  3. ^ a b Schrödinger, Erwin (1915), "Zur Theorie der Fall- und Steigversuche an Teilchen mit Brownscher Bewegung" [On the Theory of Fall- and Rise Experiments on Particles with Brownian Motion], Physikalische Zeitschrift (in German), 16 (16): 289–295
  4. ^ a b Smoluchowski, Marian (1915), "Notiz über die Berechnung der Brownschen Molekularbewegung bei der Ehrenhaft-Millikanschen Versuchsanordnung" [Note on the Calculation of Brownian Molecular Motion in the Ehrenhaft-Millikan Experimental Set-up], Physikalische Zeitschrift (in German), 16 (17/18): 318–321
  5. ^ a b Folks, J. Leroy; Chhikara, Raj S. (1978), "The Inverse Gaussian Distribution and Its Statistical Application—A Review", Journal of the Royal Statistical Society, Series B (Methodological), 40 (3): 263–275, doi:10.1111/j.2517-6161.1978.tb01039.x, JSTOR 2984691, S2CID 125337421
  6. ^ a b Bachelier, Louis (1900), "Théorie de la spéculation" [The Theory of Speculation] (PDF), Ann. Sci. Éc. Norm. Supér. (in French), Serie 3, 17: 21–89, doi:10.24033/asens.476
  7. ^ a b Bachelier, Louis (1900), "The Theory of Speculation", Ann. Sci. Éc. Norm. Supér., Serie 3, 17: 21–89 (Engl. translation by David R. May, 2011), doi:10.24033/asens.476
  8. ^ Michael, John R.; Schucany, William R.; Haas, Roy W. (1976), "Generating Random Variates Using Transformations with Multiple Roots", The American Statistician, 30 (2): 88–90, doi:10.1080/00031305.1976.10479147, JSTOR 2683801
  9. ^ Shuster, J. (1968). "On the inverse Gaussian distribution function". Journal of the American Statistical Association. 63 (4): 1514–1516. doi:10.1080/01621459.1968.10480942.
  10. ^ Schwarz, Wolfgang (2001), "The ex-Wald distribution as a descriptive model of response times", Behavior Research Methods, Instruments, and Computers, 33 (4): 457–469, doi:10.3758/bf03195403, PMID 11816448
  11. ^ Palmer, E. M.; Horowitz, T. S.; Torralba, A.; Wolfe, J. M. (2011). "What are the shapes of response time distributions in visual search?". Journal of Experimental Psychology: Human Perception and Performance. 37 (1): 58–71. doi:10.1037/a0020747. PMC 3062635. PMID 21090905.
  12. ^ Hadwiger, H. (1940). "Eine analytische Reproduktionsfunktion für biologische Gesamtheiten". Skandinavisk Aktuarietidskrijt. 7 (3–4): 101–113. doi:10.1080/03461238.1940.10404802.
  13. ^ Wald, Abraham (1944), "On Cumulative Sums of Random Variables", Annals of Mathematical Statistics, 15 (3): 283–296, doi:10.1214/aoms/1177731235, JSTOR 2236250
  14. ^ Tweedie, M. C. K. (1945). "Inverse Statistical Variates". Nature. 155 (3937): 453. Bibcode:1945Natur.155..453T. doi:10.1038/155453a0. S2CID 4113244.
  15. ^ Tweedie, M. C. K. (1956). "Some Statistical Properties of Inverse Gaussian Distributions". Virginia Journal of Science. New Series. 7 (3): 160–165.
  16. ^ Tweedie, M. C. K. (1957). "Statistical Properties of Inverse Gaussian Distributions I". Annals of Mathematical Statistics. 28 (2): 362–377. doi:10.1214/aoms/1177706964. JSTOR 2237158.
  17. ^ Tweedie, M. C. K. (1957). "Statistical Properties of Inverse Gaussian Distributions II". Annals of Mathematical Statistics. 28 (3): 696–705. doi:10.1214/aoms/1177706881. JSTOR 2237229.
  18. ^ Capacity per unit cost-achieving input distribution of rated-inverse gaussian biological neuron M Nasiraee, HM Kordy, J Kazemitabar IEEE Transactions on Communications 70 (6), 3788-3803
  19. ^ Giner, Göknur; Smyth, Gordon (August 2016). "statmod: Probability Calculations for the Inverse Gaussian Distribution". The R Journal. 8 (1): 339–351. arXiv:1603.06687. doi:10.32614/RJ-2016-024.
  20. ^ Lindsey, James (2013-09-09). "rmutil: Utilities for Nonlinear Regression and Repeated Measurements Models".
  21. ^ Swihart, Bruce; Lindsey, James (2019-03-04). "rmutil: Utilities for Nonlinear Regression and Repeated Measurements Models".
  22. ^ Wheeler, Robert (2016-09-23). "SuppDists: Supplementary Distributions".
  23. ^ Pouzat, Christophe (2015-02-19). "STAR: Spike Train Analysis with R".
  24. ^ Gjessing, Hakon K. (2014-03-29). "Threshold regression that fits the (randomized drift) inverse Gaussian distribution to survival data".
  25. ^ Hall, Byron; Hall, Martina; Statisticat, LLC; Brown, Eric; Hermanson, Richard; Charpentier, Emmanuel; Heck, Daniel; Laurent, Stephane; Gronau, Quentin F.; Singmann, Henrik (2014-03-29). "LaplacesDemon: Complete Environment for Bayesian Inference".
  26. ^ Giner, Göknur; Smyth, Gordon (2017-06-18). "statmod: Statistical Modeling".

Further reading

edit
edit