See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/275885986

Approximations to Standard Normal Distribution Function

Article  in  International Journal of Scientific and Engineering Research · April 2015

CITATIONS READS
3 634

2 authors:

Ramu Yerukala Naveen Kumar Boiroju

10 PUBLICATIONS   24 CITATIONS   
NIN, Hyderabad
63 PUBLICATIONS   70 CITATIONS   
SEE PROFILE
SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Prediction of risk of NCDs using Anthropometric data View project

I have been working on approximation of distribution function View project

All content following this page was uploaded by Naveen Kumar Boiroju on 05 May 2015.

The user has requested enhancement of the downloaded file.

International Journal of Scientific & Engineering Research, Volume 6, Issue 4, April-2015 515
ISSN 2229-5518

Approximations to Standard Normal Distribution

Function
Ramu Yerukala and Naveen Kumar Boiroju
Abstract:
This paper presents three new approximations to the cumulative distribution function of standard normal distribution. The accuracy of the proposed
approximations evaluated using maximum absolute error and the same is compared with the existing approximations available in the literature. The
proposed approximations assure minimum of three decimal value accuracy and are simple to use and easily programmable.

Keywords: Normal distribution, Maximum absolute error and Box-plots.

1. Introduction 2. Hart (1957):

The most widely used probability distribution in  −z2 
statistical applications is the normal or Gaussian 1  e π 

distribution function. The cumulative distribution function Φ 2 (z ) ≈  − 0.4 z 
2π  z + 0.8e
(cdf) of standard normal distribution is denoted by Φ (z ) 
 
and is given by
z (
exp − x 2 / 2 )dx 3. Tocher (1963):
Φ (z ) = P(Z ≤ z ) =
∫
−∞2π Φ 3 (z ) ≈
e2k z
, where k =
2
.
The cdf of normal distribution mainly used for 1+ e 2k z π
computing the area under normal curve and approximating
the t, Chi-square, F and other statistical distributions for 4. Zelen and Severo (1964):

IJSER
large samples. The cdf of normal distribution does not have  −z2 
 
a closed form. For this reason, a lot of works have been on
the development of approximations and bounds for the cdf
(
Φ 4 (z ) ≈ 1 − a1 t − a2 t 2 + a3 t 3  )e 2
2π


of normal distribution. Approximations are commonly  
 
used in many applications, where exact solutions are
numerically involved, not tractable or for simplicity and where t = (1 + 0.33267 z )−1 , a1 = 0.4361836 ,
convenience (Krishnamoorthy, 2014). a2 = 0.1201676 and a3 = 0.937298 .
There are number of approximations for computing the 5. Hart (1966):
cumulative probabilities of standard normal distribution at  
arbitrary level of accuracy available in the literature. Some  
−z2
 1 + bz 2 
of these approximations were previously studied by Polya 2
 
Φ 5 (z ) ≈ 1 −
e
1− 1 + az 2
(1945), Hart (1957 & 1966), Tocher (1963), Zelen and Severo  
2π z  −z2 
(1964), Page (1977), Hammakar (1978), Lin (1989 & 1990), 1 + bz 2
Norton (1989), Waissi and Rossin (1996), Byrc (2002)  P0 z + P0 z + e 2
2 2

 1 + az 2 
Aludaat and Alodat (2008), Winitzki (2008), Yerukala et al.
(2011) and Choudhury (2014). We propose three new 1 + 1 − 2π 2 + 6π
where a = , b = 2πa 2 and
approximation functions to approximate the cdf of 2π
standard normal distribution. The accuracy of each
π
approximation was assessed in terms of its maximum P0 = .
absolute error (Max. AE) when compared with the 2
NORMSDIST () function for the values of 0 ≤ Z ≤ 5 .
6. Page (1977):
2. Approximations to CDF of Normal Distribution Φ 6 (z ) ≈ 0.5{1 + tanh ( y )}
This section presents the historical review of different
approximations to CDF of normal distribution available in where y =
2
π
(
z 1 + 0.044715 z 2 . )
the literature for positive values of z.
7. Hammakar (1978):
1. Polya (1945):  
 2 
0.5

Φ 7 (z ) ≈ 1 − 0.51 − 1 − e − y  
 1 
   

  −2 z 2 2 
1    where y = 0.806 z (1 − 0.018 z ) .
Φ1 (z ) ≈ 1 + 1 − e π  
2   
   
 

IJSER © 2015
http://www.ijser.org
International Journal of Scientific & Engineering Research, Volume 6, Issue 4, April-2015 516
ISSN 2229-5518
8. Lin (1989): where H11 = tanh (1.280022196 − 0.720528073 z ) ,
 
Φ8 (z ) ≈ 1 − 0.5 e− 0.717 z − 0.416 z H12 = tanh(0.033142223 − 0.682842425 z ) .
2

  18. Choudhury (2014):
9. Norton (1989): −z2
  z 2 +1.2 z 0.8  e 2
Φ18 (z ) ≈ 1 −
1
 − 
 
1 − 0.5e 
2
 ; 0 ≤ z ≤ 2.7 2π 0.226 + 0.64 z + 0.33 z 2 + 3
Φ 9 (z ) ≈  The accuracy of the above functions discussed in the section
z2
 1 − 4.
 e 2 ; z > 2.7
 2π z
3. New Approximations to CDF of Normal
Distribution
10. Lin (1990): In this section, we present three new approximations to
Φ10 (z ) ≈ 1 −
1 cdf of normal distribution. A new formula for standard
1+ ey normal distribution function Φ (z ) is obtained using neural
where y = 4.2π 
 z  networks. Since Φ (z ) is symmetric about zero and
, 0 ≤ z < 9.
9 − z  Φ (− z ) = 1 − Φ (z ) . It is sufficient to approximate only for all
11. Waissi and Rossin (1996): the values of z ≥ 0 . Hence, the proposed approximations
Φ11 (z ) ≈
1 are given only for the non-negative values of z. The
( (
1 + exp − π 0.9 z + 0.0418198 z 3 − 0.0004406 z 5 )) following approximation developed using neural networks
methodology (Yerukala, 2012). A neural networks model
12. Byrc (2002A): with an input layer consisting single input node, a hidden

IJSER
layer and an output layer with one node is considered.
−z2
Φ12 (z ) ≈
(4 − π )z + 2π (π − 2)
e 2
Input node takes the values of z from zero to five with an
(4 − π )z 2 2π + 2π z + 2 2π (π − 2) increment of 0.001, whereas, the output node represents the
13. Byrc (2002B): corresponding cumulative probability for a given z. The
network is trained using backpropagation algorithm by
−z2
z 2 + a1 z + a2 taking the pair of observations as z value and its
Φ13 (z ) ≈ e 2 cumulative probability value p computed using
2π z 3 + b1 z 2 + b2 z + 2a2
NORMSDIST () function of MS Excel software. The hidden
where a1 = 5.575192695 , a2 = 12.77436324 ,
neurons are adjusted according to the sufficiently low
b1 = 14.38718147 and b2 = 31.53531977 . training and testing error. The resulting approximation to
14. Aludaat and Alodat (2008): cdf of standard normal distribution using neural networks
π is given below.
− z2
Φ14 (z ) ≈ 0.5 + 0.5 1 − e 8
Φ19 (z ) ≈
1
; 0≤ z≤5
1 + e− y
15. Winitzki (2008): where y = 0.125 + 3.611H1 − 4.658H 2 + 4.982 H 3 ,
 1 H1 = tanh (0.043 + 0.2624 z ) , H 2 = tanh (− 1.687 − 0.519 z ) and
    z 2  4 z 2   2 
H 3 = tanh (− 1.654 + 0.5044 z ) .
   −   + 0.147   
1    2  π 2    Recently, researchers have mainly concentrated on
Φ15 (z ) ≈ 1 + 1 − exp   2   developing different approximations both computationally
2   z  
1 + 0.147 tractable and sufficiently accurate by combining two or
   2  
     more existing approximations (Choudhury et al. 2007). We
  propose a combined approximation of two existing
16. Yerukala et al. (2011): approximations F1 and F2 in the form of wF1 + (1 − w)F2
0.5 − 1.136 H1 + 2.47 H 2 − 3.013H 3 ; 0 ≤ z ≤ 3.36 where the weight w is determined using the least squares
Φ16 (z ) ≈ 
1 ; z > 3.36 method over the range of z. The proposed approximation
where H1 = tanh (− 0.2695 z ) , H 2 = tanh(0.5416 z ) and has the form,
Φ 20 (z ) ≈ w Φ 5 + (1 − w)Φ13 ; z > 0 and w = 0.268 .
H 3 = tanh(0.4134 z ) .
Third approximation is a simple modification to the
17. Yerukala (2012):
Choudhury (2014) approximation and it has the form
0.46375418 + 0.065687194 H11

Φ17 (z ) ≈ − 0.602383931H12 ;0 ≤ z ≤ 3.6 Φ 21 (z ) ≈ 1 −
(
exp − z 2 / 2 )
1  44 8 5 2 
; z > 3.6  + z+ z +3
  79 5 6 
IJSER © 2015
http://www.ijser.org
International Journal of Scientific & Engineering Research, Volume 6, Issue 4, April-2015 517
ISSN 2229-5518
Accuracy of the proposed approximations discussed in Φ11 (z ) 4.37E-05 1.14 and 1.15
the following section.
Φ12 (z ) 7.18E-04 1.07 to 1.12
4. Results and Discussion
This section presents the approximation errors of the Φ13 (z ) 1.87E-05 1.47 to 1.57
listed functions in the above two sections. Each Φ14 (z ) 1.97E-03 1.83 to 1.93
approximation is evaluated on the basis of maximum
absolute error from NORMSDIST() function within the Φ15 (z ) 6.20E-05 2.19 to 2.23
given range of z values. Table 1 presents the maximum
absolute error and corresponding z value of each of the Φ16 (z ) 1.25E-03 2.58 to 2.69
approximation. From this table, it is clear that the proposed
Φ17 (z ) 1.17E-04 3.67 and 3.68
approximation Φ 20 (z ) is more efficient than all the listed
approximations and Φ19 (z ) and Φ 21 (z ) are very good Φ18 (z ) 1.93E-04 0.00
alternatives to the existing approximations. The error Φ19 (z ) 1.61E-04 0.00
distribution of the approximations shows that, the
proposed approximations have the error very close to zero Φ 20 (z ) 7.54E-06 1.89 to 1.91
(Figure 1). Maximum absolute error for Φ19 (z ) and Φ 21 (z )
Φ 21 (z ) 1.10E-04 0.06
is observed at around the origin, whereas the same is
observed for Φ 20 (z ) at z=1.9. The approximation Φ 20 (z )
have reduced the error about 86% and 60% respectively as
compared with the Φ 5 (z ) and Φ13 (z ) . Approximation
Φ 20 (z ) provides 5 to 9 correct decimals within the given

IJSER
range of z (Figure 2). The approximation Φ 21 (z ) is derived
by improving the coefficients of the function proposed by
Choudhury (2014), the resulting approximation not only
improved the approximation accuracy but also simplified
the formula at a great extent.
Finally, the proposed approximations are very simple
and provides a minimum accuracy of three decimal places.
Simplicity of these approximations enables their application
over a wide range of analytical studies at a reasonable
accuracy levels. These approximations are easy to
programme and simple to use on a handheld calculator.
Figure 1. Error Distribution of the Approximations
Table 1. Maximum absolute error and corresponding z
value of the approximations
Function Max. A E Occurrence of Max. AE at Z=z

Φ1 (z ) 3.15E-03 1.64 to 1.66

Φ 2 (z ) 4.30E-03 0.28 to 0.31

Φ 3 (z ) 1.77E-02 1.7 to 1.77

Φ 4 (z ) 1.15E-05 0.51 to 0.54

Φ 5 (z ) 5.32E-05 1.02 to 1.06

Φ 6 (z ) 1.79E-04 1.22 to 1.27 and 2.56 to 2.62

Φ 7 (z ) 6.23E-04 0.33 and 0.34 Figure 2. Error distribution of the approximations Φ 5 (z ) ,

Φ8 (z ) 6.59E-03 0.39 Φ13 (z ) and Φ 20 (z )

Φ 9 (z ) 8.07E-03 0.87 to 0.89 Acknowledgements:

We are very thankful to Dr. M. Krishna Reddy,
Φ10 (z ) 6.69E-03 0.44 and 0.45
Professor (Rtd.), Department of Statistics, Osmania
IJSER © 2015
http://www.ijser.org
International Journal of Scientific & Engineering Research, Volume 6, Issue 4, April-2015 518
ISSN 2229-5518
University, Hyderabad for his valuable suggestions and Function,” Applied Mathematical Sciences, Vol. 2,
encouragement. no.9, 425-429, 2008.
13. K.. Krishnamoorthy, “Modified Normal-Based
References: Approximation to the Percentiles of Linear
1. A. Choudhury and P. Roy, “A Fairly Accurate Combination of Independent Random Variables
Approximation to the Area under Normal Curve,” with Applications,” Available online at
Communications in Statistics-Simulation and http://www.ucs.louisiana.edu/~kxk4695/CS-
Computation, 38, pp. 1485-1492, 2009. MNA-R.pdf, 2014
2. A. Choudhury, “A Simple Approximation to the 14. M. Zelen and N. C. Severo, “Probability Function
Area under Standard Normal Curve”, Mathematics (925-995) in Handbook of Mathematical
and Statistics 2(3)147-149, 2014. Functions”, Edited by M. Abramowitz and I. A.
3. A. Choudhury, S. Ray, and P. Sarkar, Stegun, Applied Mathematics Series, 55, 1964.
“Approximating the Cumulative Distribution 15. R. G. Hart, “A close approximation related to the
Function of the Normal Distribution,” Journal of error function (in Technical Notes and Short
Statistical Research, 41(1), 59–67, 2007. Paper),” Math. Comput. 20 (96) 600, 1966.
4. B.I. Yun, "Approximation to the Cumulative 16. R. G. Hart, “A formula for the approximation of
Normal Distribution using Hyperbolic Tangent definite integrals of the normal distribution
based Function," J. Korean Math. Soc. 46, No. 6, pp. function,” Mathematical Tables and other Aids to
1267-1276, 2009. Computation 11(60), 1957.
5. E. Page, “Approximating to the Cumulative 17. R. M. Norton, “Pocket-Calculator Approximation
Normal function and its Inverse for use on a Pocket for Area under the Standard Normal Curve,” the
Calculator,” Applied Statistics, 26, 75-76, 1977. American Statistician, Vol. 43, No. 1, 1989.
6. G. Polya, “Remarks on computing the probability 18. R. Yerukala, “Functional Approximation using

IJSER
integral in one and two dimensions”, 1st Berkeley Neural Networks,” Unpublished thesis,
Symp. Math. Statist. Prob., pp. 63-78, 1945. Department of Statistics, Osmania University,
7. G. R. Waissi, and D. F. Rossin, "A Sigmoid Hyderabad, Telangana, India, 2012.
Approximation of the Standard Normal Integral," 19. R. Yerukala, N. K., Boiroju and M. K. Reddy, “An
Applied Mathematics and Computation 77, pp 91-95, Approximation to the CDF of Standard Normal
1996. Distribution,” International Journal of Mathematical
8. H. C. Hammakar, “Approximating the Cumulative Archive 2(7):1077-1079, 2011.
Normal Distribution and its Inverse,” Applied 20. S. Winitzki, “A Handy approximation for the Error
Statistics, 27, 76-77, 1978. Function and its Inverse,” Available online at
9. J. T. Lin, “A Simpler Logistic Approximation to the http://sites.google.com/site/winitzki/sergei-
Normal Tail Probability and its Inverse”, Appl. winitzkis-files, 2008 .
Statist. 39, No.2, pp. 255-257, 1990. 21. W. Bryc, “A Uniform Approximation to the Right
10. J. T. Lin, “Approximating the Normal Tail Normal Tail Integral”, Applied Mathematics and
Probability and its Inverse for use on a Pocket- Computation 127, pp 365-374, 2002.
Calculator,” Applied Statistics, 38, 69-70, 1989.
11. K. D. Tocher, “The Art of Simulation”, English
University Press, London, 1963.
12. K. M. Aludaat and M. T. Alodat, “A Note on
Approximating the Normal Distribution Ramu Yerukala, ANURAG Group of Institutions,
Hyderabad, Telangana, India-500 088.
Email: [email protected]

IJSER © 2015
http://www.ijser.org

View publication stats