REVSTAT Statistical Journal

Volume 10, Number 2, June 2012, 211227

COMPARISON OF CONFIDENCE INTERVALS FOR

THE POISSON MEAN: SOME NEW ASPECTS
Authors:

V.V. Patil
Department of Statistics, Tuljaram Chaturchand College,
Baramati, Maharashtra, India
vaishu_kop@indiatimes.com

H.V. Kulkarni
Department of Statistics, Shivaji University,
Kolhapur, Maharashtra, India
kulkarnih1@rediffmail.com

Received: February 2010

Revised: May 2011

Accepted: January 2012

Abstract:
We perform a comparative study among nineteen methods of interval estimation of
the Poisson mean, in the intervals (0,2), [2,4] and (4,50], using as criteria coverage, expected length of confidence intervals, balance of noncoverage probabilities, E(P-bias)
and E(P-confidence). The study leads to recommendations regarding the use of particular methods depending on the demands of a particular statistical investigation and
prior judgment regarding the parameter value if any.

Key-Words:
confidence intervals (CI); Poisson mean; comparison.

AMS Subject Classification:

62-07, 62F25.

212

V.V. Patil and H.V. Kulkarni

Comparison of Confidence Intervals for the Poisson Mean: Some New Aspects

1.

213

INTRODUCTION

Construction of CIs in discrete distributions is a widely addressed problem.

The standard method of obtaining a 100 (1 )% CI for the Poisson mean is
based on inverting an equal tailed test for the null hypothesis H0 : = 0 . This
is an exact CI, in the sense that it is constructed using the exact distribution.
Exact CIs are very conservative and too wide. A large number of alternate
methods for obtaining CIs for based on approximations for the Poisson distribution are suggested in the literature to overcome these drawbacks. Desirable
properties of those approximate CIs are:
for (1 ) confidence interval the infimum over of the coverage probability should be equal to (1 );
confidence interval can not be shortened without the infimum of the
coverage falling below (1 ).
We attempt to perform an exhaustive review of the existing methods for
obtaining confidence intervals for the Poisson parameter and present an extensive
comparison among these methods based on the following criterion:
1) Expected length of confidence intervals (E(LOC)),
2) Percent coverages (Coverage),
3) E(P-bias) and E(P-confidence),
4) Balance of right and left noncoverage probabilities.
Section 2 enumerates several methods for interval estimation of , giving
appropriate references. Section 3 describes criteria used for comparison, Section 4
reports details of the comparative study and Section 5 presents concluding remarks.

2.

A REVIEW OF THE EXISTING METHODS

Table 1 presented next reports 19 CIs for the Poisson mean. In the table,
Schwertman and Martinez is abbreviated as SM, Freeman and Tukey by FT,
Wilson and Hilferty by WH, continuity correction by CC, Second Normal
by SN and Likelihood Ratio by LR. Furthermore, 1 = /2, 2 = 1 /2, and
xc = x + c for any number c.

214

V.V. Patil and H.V. Kulkarni

Table 1:

Confidence limits for the nineteen methods.

Name and reference

Lower Limit

1: Garwood (GW)
(1936)
2: WH (WH)
(1931)

2(2x,1 )

x + Z1

4: SN (SN)
SM (1994)

2
x + Z
/2 + Z1
1

5: Wald CC (FNCC)
SM (1994)

7: Molenaar (MOL)
(1970)

x0.5 + Z1

9: Vandenbroucke (SR)
(1982)

11: FT (FT)
(1950)
12: Hald (H)
(1952)

3
x

x + Z2

q
2 /4
x + Z
1

0.25

x0.5

x0.5 + Z2

2
xc + Z1/2

x1 + Z1

2


1

x1

q
2 /4
x + Z
2

x0.5

2
x + Z2/2

2
xc + Z2 /2

p

2
x + 3/8 + Z2 /2 3/8

0.25

2
x.02 + Z1/2

x+

x1 + Z2

2


1

2
x.5 + Z2 /2 + .5

2
x.96 + Z2 /2

For x = 0; 0
For x > 0; Wald limit

For x = 0; log(1 )
For x > 0; Wald limit

For x = 0; 0
For x > 0; Bartlett limit

For x = 0; log(1 )
For x > 0; Bartlett limit

16: LR (LR)
Brown et al. (2003)

No closed form

No closed form

17: Jeffreys (JFR)

Brown et al. (2003)

G 1 , x0.5 , 1/r

18: Mid-P
Lancaster (1961)

No closed form

15: Modified Bartlett (MB)

Barker (2002)

19: Approximate Bootstrap

Confidence (ABC)
Swift (2009)

Z0 + Z1

2 x
1 a(Z0 + Z1)

where a = Z0 = 1/(6 x )

x+

3

2
x0.5 + Z
/2 + Z2
2

.5
2
/4
x0.5 + Z
1


2
+ 1 /6 + Z2
x0.5 + 2 Z
2



.5
2
x0.5 + Z2 + 2 /18

2
x.5 + Z1/2 + .5

2
x + Z
/2 + Z2
1

2
x + Z1/2

x+

2

x1 1 1/9 x1 + Z2/3

p

2
x + 3/8 + Z1/2 3/8

13: Begaud (BB)

(2005)
14: Modified Wald (MW)
Barker (2002)

2(2x1 ,2 )

2
x0.5 + Z
/2 + Z1
1

.5
2
/4
x0.5 + Z
2


2
+ 1 /6 + Z1
x0.5 + 2 Z
1



.5
2
x0.5 + Z1 + 2 /18

8: Bartlett (BART)
(1936)

10: Anscombe (ANS)

(1948)

2

x 1 1/9 x + Z1/3

3: Wald (W)
SM (1994)

6: SN CC (SNCC)
SM (1994)

Upper Limit

G 2 , x0.5 , 1/r

No closed form
x+

Z0 + Z2
1 a(Z0 + Z2 )

2

Comparison of Confidence Intervals for the Poisson Mean: Some New Aspects

3.

215

CRITERIA FOR COMPARISON

The criteria considered for the comparison among the above mentioned CIs
are E(LOC) of CIs, coverage probability, ratio of the left to right noncoverage
probabilities, E(P-confidence) and E(P-bias).
Here we explain the details of the three criterion for comparison mentioned
in Section 1. Without loss of generality a sample of size n = 1 is considered. The
comparisons are carried out over (0, 50].
P
The expected value of a function g(x) is computed as
x=0 g(x) p (x) where

x
p (x) = e /x!. The infinite sums in the computation of these quantities were
approximated by appropriate finite ones up to 0.001 margin of error.
The coverage probability C(), noncoverage probability on the left L(),
noncoverage probability on the right R(), and corresponding expected length


E(LOC) of a CI l(x), u(x) are respectively computed by taking g(x) = I l(x)








u(x) , I > u(x) , I < l(x) and u(x) l(x) , where I() is the indicator
function of the bracketed event.

3.1. Computation of E(P-confidence) and E(P-bias)

Let CI(x) be the CI obtained for the observation x having nominal level (1
)100%. The P-bias and P-confidence are defined in terms of the standard equal


tailed P-value function P (, x) = min 2 P (X x), 2 P (X x), 1 . The P-confidence of the CI that measures how strongly the observation x rejects parameter




P (, x) 100%.
values outside CI is defined as Cp CI(x), x = 1 supCI(x)
/
The P-bias of a CI which quantifies the largeness of P-values for values of outside the CI in comparison with those inside the CI is given by




b CI(x), x = max 0, supCI(x)
P (, x) inf CI(x) P (, x) 100%. For the
/
Poisson distribution P (, x) is continuous and a monotone function in in opposite directions to the left and right of the interval for each value of x. Hence
the supremums and infimums occur at the upper or lower end points of the CIs.
Consequently the formulae of P-bias and P-confidence are reduced to

n





o
Cp CI(x), x = 1 max 2P X x; = l(x) , 2P X x; = u(x)
100% ,
 n






o
b CI(x), x = max 0, 2P X x; = l(x) 2P X x; = u(x)
100% .
Their expected values are computed as described above.

216

V.V. Patil and H.V. Kulkarni

It was observed that when the actual value of is a fraction, the CI with
their endpoints rounded to the nearest integer (for lower limit, rounding to an
integer less than the limit and reverse for the upper limit) improved coverage
probabilities to a very large extent at the cost of increasing E(LOC) at most
by one unit. This is clearly visible from Figure 1 which displays the Box plot
of coverages for the rounded and unrounded CIs obtained using Wald method.
Similar pattern was observed for other methods.

Coverage

95

90

85
unrounded

rounded

Figure 1: Impact of rounding on coverage of Wald CI.

Consequently the E(LOC) and percent coverages reported here correspond

to these rounded intervals and the comparison carried out among the methods in
the sequel is based on rounded intervals.

4.

COMPARISON AMONG THE METHODS

4.1. Comparison based on coverages and E(LOC)

On careful examination revealed that different methods perform differently

in certain subsets of the parameter space.
Consequently the performance of each method was studied separately on
the three regions, namely (0,2), [2,4] and (4,50] in the parameter space. Panels
(a) and (b) of Figures 2A to 4A display respectively the boxplots of coverages and
graphs of relative E(LOC) of conservative methods (i.e. ratio of E(LOC) for the
concerned method to the same for Garwood exact CI) for different regions defined
above. Figures 2B to 4B display similar plots for nonconservative methods.
The observations from these graphs are tabulated in Table 2. The methods
displayed in bold face have shortest length among the concerned group. Here
G1 = {GW, MOL, WH, BB} and G2 = {BART, W, H}.

217

Comparison of Confidence Intervals for the Poisson Mean: Some New Aspects

Table 2:

Coverage performance of the nineteen methods.

Type

(0,2)

[2,4]

(4,50]

Conservative

FNCC, LR, ANS, G1,

FT, JFR, MB, MW, SN
SNCC, ABC, SR, Mid-P

MB, ANS, SN, G1

ABC, SR, JFR
SNCC, Mid-P

G1, SNCC, ABC, LR

H, BART, MW, ANS
FT, MB, SN, Mid-P
JFR, FNCC, W, SR

NonConservative

G2

G2, FNCC, FT, LR

MW

SNCC
MOL,GW,
WH,BB
ANS

100

2
1.8
1.6
1.4
1.2
1
0.8
0.6
0.4
0.2
0

99

Relative lengths

98

Cov

97
96
95
94
93
92
FNCC MW ANS ABC

FT

G1

JFR MB SNCC SR

LR Mid-P SN

MB
JFR
LR
MW
FT
FNCC

ABC

Methods
SR

mu

(a) Boxplot of coverages for conservative methods.

(b) Relative lengths of conservative methods.

Figure 2A: Coverages and relative E(LOC) for conservative methods for
parametric space (0,2), where G1 = {GW, MOL, WH, BB}.

1.6
100

1.4
Relative length

90
80
70

Cov

60
50
40
30

1.2

SN

0.8

BART

0.6

0.4

20

0.2

10

midp

0
BART

Methods

(a) Boxplot of coverages for nonconservative methods.

mu

(b) Relative lengths of nonconservative

methods.

Figure 2B: Coverages and relative E(LOC) for nonconservative methods

for parametric space (0,2).

218

V.V. Patil and H.V. Kulkarni

Relative length

100

Cov

95

90

1.4

SNCC

1.2

G1

SN

0.8

ANS

0.6
MB

0.4
JFR

0.2
ABC

0
2

85
ABC

G1

MB

SNCC ANS

SN

JFR

SR

2.5

Mid-p

3.5

SR

mu

Methods

(a) Boxplot of coverages for conservative methods.

(b) Relative lengths of conservative methods.

Figure 3A: Coverages and relative E(LOC) for conservative methods for
parametric space [2,4], where G1 = {GW, MOL, WH, BB}.

1.2

100

Relative length

Cov

90

FNCC

0.8

BART
FT

0.6

0.4

MW
LR

0.2

midp

80

2
BART FNCC

FT

LR

MW

2.5

3.5

Methods

(a) Boxplot of coverages for nonconservative methods.

mu

(b) Relative lengths of nonconservative

methods.

Figure 3B: Coverages and relative E(LOC) for nonconservative methods

for parametric space [2,4].

219

Comparison of Confidence Intervals for the Poisson Mean: Some New Aspects

99

1.09
1.08
1.07
1.06
1.05
1.04
1.03
1.02
1.01
1.00
0.99

Relative length

Cov

98

97

96

SNCC

G1

95
G1

14

24

34

44

mu

SNCC

Methods

(a) Boxplot of coverages for conservative methods.

(b) Relative lengths of conservative methods.

Figure 4A: Coverages and relative E(LOC) for conservative methods for
parametric space (4,50], where G1 = {GW, MOL, WH, BB}.

100

1.05

Relative length

1
95

Cov

0.95

90

0.9

0.85
0.8
0

ABC ANS BARTFNCC FT

H JFR LR Mid-p MB

MW

SN SR

10

20

30

40

50

W
SN
FNCC
BART
ANS
FT
H
MW
MB
JFR
LR
Mid-p
ABC
SR

mu

Methods

(a) Boxplot of coverages for nonconservative methods.

(b) Relative lengths of nonconservative

methods.

Figure 4B: Coverages and relative E(LOC) for nonconservative methods

for parametric space (4,50].

220

V.V. Patil and H.V. Kulkarni

4.2. Comparison with respect to balance of noncoverage probabilities

For a two sided CI procedure it is desirable to have the right and left noncoverage probabilities to be fairly balanced. We plot the ratio of the left to right
noncoverage probabilities as a function of Poisson mean for the nineteen methods
in Figure 5A and 5B for regions (2,4) and (4,50). For balanced noncoverage, ratio
should oscillate in the close neighborhood of 1. For region (0,2) all methods are
well below 1, with the exception of Wald method.
A careful observation of figures leads to the following region wise performance of methods with respect to right-to-left noncoverage balance reported in
Table 3.

Table 3:

Performance based on right-to-left noncoverage balance.

Performance

(2,4)

(4,50)

Fairly balanced around 1

G1, ABC, LR, JFR, SR

Uniformly below 1

SNCC, SN, G1, ABC, MB

SN, SNCC, Mid-P

Uniformly above 1

LR, JFR, SR, Mid-P

FT, ANS, FNCC, MW, G2

FT, MB, ANS, FNCC, MW, G2

4.3. Comparison based on E(P-bias) and E(P-confidence)

For comparison of methods on the basis of E(P-bias) and E(P-confidence),

we consider three regions of sample space (0,2), (2,4) and (4,50). Three panels
(a) to (c) of Figures 6 and 7 represent boxplots of E(P-confidence) and E(P-bias)
for these three regions. Recommendations on the basis of E(P-bias) and E(P-confidence) for two regions tabulated in Table 4.

Table 4:

Recommendations on the basis of E(P-bias) and E(P-confidence).

Performance

(0,2)

(2,4)

(4,50)

Smallest E(P-bias)
Largest E(P-confidence)

FNCC, MW, W
SNCC, SN, ABC, G1

FNCC, SNCC
SNCC, SN, G1

SNCC, Mid-P, SN
SNCC, G1, SN

221

Comparison of Confidence Intervals for the Poisson Mean: Some New Aspects

JFR
15
10

L/R

L/R

Mid-P
6
4
2
0

5
0

mu

LR

15
10

L/R

L/R

SR

5
0
0

30
20
10
0

mu

mu

ANS

FT
100
L/R

50
L/R

mu

25

50

0
0

mu

BART
100
L/R

100
L/R

3
mu

50

50
0

0
2

mu

mu

MW

FNCC

600
500

L/R

L/R

400
300
200
100
0

2000
1500
1000
500
0
2

mu

mu

L/R

W
800
600
400
200
0
2

mu

Figure 5A: Graph of ratio of noncoverage probabilities for parametric space (2,4].
The ratio of noncoverage probabilities for methods SNCC, SN, G1,
ABC and MB are zero for parametric space (2,4].

222

V.V. Patil and H.V. Kulkarni

SN
2

0.8

1.5

0.6

L/R

L/R

SNCC
1

0.4

1
0.5

0.2

0
0

20

40

60

20

40

ABC

1.5

1.5
L/R

L/R

LR

0.5

0.5

0
0

20

40

60

20

mu

40

60

mu

SR

JFR

2.5

2.5

1.5

1.5

L/R

L/R

60

mu

mu

1
0.5

1
0.5

0
0

20

40

60

20

mu

40

60

mu

Mid-P

G1
6
5
4

4
L/R

L/R

3
2

3
2
1

1
0

10

20

30
mu

40

50

60

20

40

60

mu

(continues)

223

Comparison of Confidence Intervals for the Poisson Mean: Some New Aspects

(continued)

FT

ANS
6

5
4
L/R

L/R

4
3

3
2
1

2
1

0
0

20

40

60

20

BART

60

40

60

MB

30

30

25

25

20

20
L/R

L/R

40
mu

mu

15

15

10

10

5
0

0
0

20

40

60

20
mu

mu

MW

30

80

25
60
L/R

L/R

20
15

40

10
20

5
0

20

40

60

20

mu

60

40

60

FNCC

80

300

60

250
200
L/R

L/R

40
mu

40

150
100
50

20
0

0
0

20

40
mu

60

20
mu

Figure 5B: Graph of ratio of non coverage probabilities for parametric space (4,50].

224

V.V. Patil and H.V. Kulkarni

100
90

E(P-confidence)

80
70
60
50
40
30
20
10
0
SNCCABC G1 SN JFRSR ANS MB FTMid-P FNCCLRMWBART H

Methods

Figure 6(a): Boxplot of E(P-confidence) for parametric space (0,2].

E(P-confidence)

100

90

80

70

SNCC G1 SN ABC SR J FRFT ANS H MB FNCCMid-pBARTLR MW W

Methods

Figure 6(b): Boxplot of E(P-confidence) for parametric space (2,4].

E(P-confidence)

100

90

80
SNCC G1 SN Mid-p ABC SR JFR FT ANSBART H

LR MBFNCC MW W

Methods

Figure 6(c): Boxplot of E(P-confidence) for parametric space (4,50).

Comparison of Confidence Intervals for the Poisson Mean: Some New Aspects

E(P-bias)

1.5

1.0

0.5

0.0
FNCCMW W ANSSNCC G1 ABC SR

FT JFRBART H

LR MB SN Mid-p

Methods

Figure 7(a): Boxplot of E(P-bias) for parametric space (0,2].

E(P-bias)

5
4
3
2
1
0
FNCC SNCC G1ANS MW W ABC SR FT J FRSN Mid-PBART H

LR MB

Methods

Figure 7(b): Boxplot of E(P-bias) for parametric space (2,4].

E(P-bias)

10

0
SNCCMid-PSN G1 ABC J FR SR ANS FT LR H MBBARTFNCCMW W

Methods

Figure 7(c): Boxplot of E(P-bias) for parametric space (4,50).

225

226

5.

V.V. Patil and H.V. Kulkarni

CONCLUDING REMARKS

Rounding of end points of CI considerably improves the coverages of CI.

Our remarks are based on rounded intervals. A best choice for CI depends on the
objectives of the underlying investigations and a broad prior knowledge about
the underlying parameter if any.
Finally, our investigation suggests the following recommendations:
1) In the analysis of rare events where is expected to be very small
in between 0 to 2, we recommend MW and FNCC method on the
basis of highest coverage probabilities with shortest expected length
and smallest expected P-bias and reasonable expected P-confidence.
In this region LR is also recommendable on the basis of all the criteria
except E(P-bias).
2) For the situations where the parameter is expected to be large more
than 4, methods involved in G1 are the best choice. In fact the performance of methods in G1 is uniformly satisfactory (if not best) on the
entire parameter space with respect to all the criteria, so in the absence
of any knowledge regarding the underlying parameter, we recommend
these methods for use.
3) We strongly recommend to avoid using W, BART, and H methods in all
kinds of applications, since these are uniformly nonconservative for all
parameter values, have large E(P-bias) and smallest E(P-confidence)
and highly imbalanced noncoverage on the right and left side.
These recommendations are useful guidelines for consulting professionals,
in data analysis, software development, and can be an interesting addition to the
discussion of case studies in Applied Statistic courses.

ACKNOWLEDGMENTS

We are very much grateful to anonymous referees and the Associated Editor
for many suggestions that greatly improved the manuscript. H.V. Kulkarni was
supported by the grants received by Government of India, Department of Science
and Technology, India, under the project Reference No. SR/S4/MS: 306/05.

Comparison of Confidence Intervals for the Poisson Mean: Some New Aspects

227

REFERENCES

[1]

Anscombe, F.J. (1948). The transformation of Poisson, Binomial and Negative

Binomial data, Biometrika, 35, 246254.

[2]

Barker, L. (2002). A comparison of nine confidence intervals for a Poisson

parameter when the expected number of events is 5, The American Statistician,
56(2), 8689.

[3]

Bartlett, M.S. (1936). The square root transformation in the analysis of variance, Journal of Royal Statistical Society Supplement, 3, 6878.

[4]

Begaud, B.; Karin, M.; Abdelilah, A.; Pascale, T.; Nicholas, M. and
Yola, M. (2005). An easy to use method to approximate Poisson confidence
limits, European Journal of Epidemiology, 20(3), 213216.

[5]

Brown, Cai and Dasgupta (2003). Interval estimation in exponential families,

Statistica Sinica, 13, 1949.

[6]

Freeman, M.F. and Tukey, J.W. (1950). Transformations related to the angular and square root, Annals of Mathematical Statistics, 21, 606611.

[7]

Garwood, F. (1936). Fiducial limits for the Poisson distribution, Biometrika,

28, 437442.

[8]

Hald, A. (1952). Statistical Theory with Engineering Applications, John Wiley

and Sons, New York.

[9]

Lancaster, H.O. (1961). Significance tests in discrete distributions, Journal of

American Statistical Association, 56, 223234.

[10]

Molenaar, W. (1970). Approximations to the Poisson, Binomial and Hypergeometric Distribution Functions, Mathematical Center Tracts 31, Mathematisch
Centrum, Amsterdam.

[11]

Schwertman, N.C. and Martinez, R.A. (1994). Approximate Poisson confidence limits, Communication in Statistics Theory and Methods, 23(5), 1507
1529.

[12]

Swift, M.B. (2009). Comparison of confidence intervals for a Poisson MeanFurther considerations, Communication in Statistics Theory and Methods, 38,
748759.

[13]

Vandenbroucke, J.P. (1982). A shortcut method for calculating the 95 percent confidence interval of the standardized mortality ratio, (Letter), American
Journal of Epidemiology, 115, 303304.

[14]

Wilson, E.B. and Hilferty, M.M. (1931). The distribution of chi-square,

Proceedings of the National Academy of Sciences, 17, 684688.