NCSS Statistical Software NCSS.

com

Chapter 230

Circular Data
Analysis
Introduction
This procedure computes summary statistics, generates rose plots and circular histograms, computes hypothesis
tests appropriate for one, two, and several groups, and computes the circular correlation coefficient for circular
data.
Angular data, recorded in degrees or radians, is generated in a wide variety of scientific research areas. Examples
of angular (and cyclical) data include daily wind directions, ocean current directions, departure directions of
animals, direction of bone-fracture plane, and orientation of bees in a beehive after stimuli.
The usual summary statistics, such as the sample mean and standard deviation, cannot be used with angular
values. For example, consider the average of the angular values 1 and 359. The simple average is 180. But with a
little thought, we might conclude that 0 is a better answer. Because of this and other problems, a special set of
techniques have been developed for analyzing angular data. This procedure implements many of those techniques.

230-1
© NCSS, LLC. All Rights Reserved.
NCSS Statistical Software NCSS.com
Circular Data Analysis

Technical Details
Suppose a sample of n angles a1 , a2 ,..., a n is to be summarized. It is assumed that these angles are in degrees.
Fisher (1993) and Mardia & Jupp (2000) contain definitions of various summary statistics that are used for
angular data. These results will be presented next. Let
n n
Cp Sp
Cp = ∑ cos( pai ) , Cp =
i =1 n
, Sp = ∑ sin( pa ) , S
i =1
i p =
n
,

Rp
Rp = C p2 + S p2 , Rp =
n
 −1  S p 
tan   C p > 0, S p > 0
  Cp 
  Sp 

Tp = tan −1   + π Cp < 0
  Cp 
 S 
tan −1  p  + 2π S p < 0, C p > 0
  Cp 
To interpret these quantities it may be useful to imagine that each angle represents a vector of length one in the
direction of the angle. Suppose these individual vectors are arranged so that the beginning of the first vector is at
the origin, the beginning of the second vector is at the end of the first, the beginning of the third vector is at the

end of the second, and so on. We can then imagine a single vector a that will stretch from the origin to the end of
the last observation.
 
R1 , called the resultant length, is the length of a . R1 is the mean resultant length of a . Note that R1 varies
between zero and one and that a value of R1 near one implies that there was little variation in values of the
angles.
The mean direction, θ , is a measure of the mean of the individual angles. θ is estimated by T1 .
The circular variance, V, measures the variation in the angles about the mean direction. V varies from zero to one.
The formula for V is
V = 1 − R1
The circular standard deviation, v, is defined as

v = − 2 ln( R1 )

The circular dispersion, used in the calculation of confidence intervals, is defined as

1 − T2
δ=
2 R12
The skewness is defined as
R2 sin(T2 − 2T1 )
s =
(1 − R )
3/ 2
1

230-2
© NCSS, LLC. All Rights Reserved.
NCSS Statistical Software NCSS.com
Circular Data Analysis

The kurtosis is defined as

R2 cos(T2 − 2T1 ) − R14
k =
(1 − R )
2
1

Correction for Grouped Data

When the angles are grouped, a multiplicative correction for R may be necessary. The corrected value is given by
R p* = gR p
where
π/J
g =
sin(π / J )
Here J is the number of equi-sized arcs. Thus, for monthly data, J would be 12.

Confidence Interval for the Mean Direction

Upton & Fingleton (1989) page 220 give a confidence interval for the mean direction when no distributional
assumption is made as
T1 ± sin −1 ( zα / 2σ )
where

n(1 − H )
σ =
4 R2
1 n n

H = cos(2T1 )∑ cos(2ai ) + sin(2T1 )∑ sin(2ai )
n i =1 i =1 

Circular Uniform Distribution

Uniformity refers to the situation in which all values around the circle are equally likely. The probability
distribution on a circle with this property is the circular uniform distribution, or simply, the uniform distribution.
The probability density function is given by
1
f (a ) =
360
The probability between any two points is given by
a 2 − a1
Pr(a1 < a 2 | a1 ≤ a 2 , a 2 ≤ a1 + 2π ) =
360

230-3
© NCSS, LLC. All Rights Reserved.
NCSS Statistical Software NCSS.com
Circular Data Analysis

Tests of Uniformity
Uniformity refers to the situation in which all values around the circle are equally likely. Occasionally, it is useful
to perform a statistical test of whether a set of data do not follow the uniform distribution. Several tests of
uniformity have been developed. Note that when any of the following tests are rejected, we can conclude that the
data were not uniform. However, when the test is not rejected, we cannot conclude that the data follow the
uniform distribution. Rather, we do not have enough evidence to reject the null hypothesis of uniformity.
Rayleigh Test
The Rayleigh test, discussed in Mardia & Jupp (2000) pages 94-95, is the score test and the likelihood ratio test
for uniformity within the von Mises distribution family. The Rayleigh test statistic is 2nR 2 . For large samples, the
distribution of this statistic under uniformity is a chi-square with two degrees of freedom with an error of
( ) . A closer approximation to the chi-square with two degrees of freedom is achieved by
approximation of O n
−1

the modified Rayleigh test. This test, which has an error of O( n ) , is calculated as follows.
−2

 1 nR 4
S * = 1 −  2nR 2 +
 2n  2

Modified Kuiper's Test

The modified Kuiper's test, Mardia & Jupp (2000) pages 99-103, was designed to test uniformity against any
alternative. It measures the distance between the cumulative uniform distribution function and the empirical
distribution function. It is accurate for samples as small as 8. The test statistic, V, is calculated as follows

 0.24 
V = Vn  n + 0155
. + 
 n
where
a i a i 1
Vn = max  ( i ) −  − min  ( i ) −  +
i =1 to n  360 n i =1 to n  360 n  n
Published critical values of V are
V Alpha
1.537 0.150
1.620 0.100
1.747 0.050
1.862 0.025
2.001 0.010
This table was used to create an interpolation formula from which the alpha values are calculated.
Watson Test
The following uniformity test is outlined in Mardia & Jupp pages 103-105. The test is conducted by calculating
U 2 and comparing it to a table of values. If the calculated value is greater than the critical value, the null
hypothesis of uniformity is rejected. Note that the test is only valid for samples of at least eight angles.
The calculation of U 2 is as follows

i − 12
2
n
 1 1
U = ∑ u( i ) −
2
−u +  +
i =1  n 2  12n

230-4
© NCSS, LLC. All Rights Reserved.
NCSS Statistical Software NCSS.com
Circular Data Analysis

where
n

∑u (i)
a( i )
u= i =1
, u( i ) =
n 360
a(1) ≤ a( 2 ) ≤ a( 3) ≤ ≤ a( n ) are the sorted angles. Note that maximum likelihood estimates of κ and θ are used
in the distribution function. Mardia & Jupp (2000) present a table of critical values that has been entered into
NCSS. When a value of U 2 is calculated, the table is interpolated to determine its significance level.
Published critical values of U 2 are
U2 Alpha
0.131 0.150
0.152 0.100
0.187 0.050
0.221 0.025
0.267 0.010

Von Mises Distributions

The Von Mises distribution takes the role in circular statistics that is held by the normal distribution in standard
linear statistics. In fact, it is shaped like the normal distribution, except that its tails are truncated.
The probability density function is given by

f (a; θ , κ ) =
1
2π I 0 (κ )
[
exp κ cos(a − θ ) ]
where I p ( x ) (the modified Bessel function of the first kind and order p) is defined by

∞ 2r + p
1  x
I p ( x) = ∑   , p = 0,1,2,
r = 0 ( r + p ) !r ! 2
 

In particular
∞ 2r
1  x
I0 ( x ) = ∑ 2  
r = 0 ( r !)
 2
2π
1 x cos (θ )
=
2π ∫e
0
dθ

The parameter θ is the mean direction and the parameter κ is the concentration parameter.
The distribution is unimodal. It is symmetric about A. It appears as a normal distribution that is truncated at plus
and minus 180 degrees. When κ is zero, the von Mises distribution reduces to the uniform distribution. As κ gets
large, the von Mises distribution approaches the normal distribution.

230-5
© NCSS, LLC. All Rights Reserved.
NCSS Statistical Software NCSS.com
Circular Data Analysis

Point Estimation
The maximum likelihood estimate of θ is the sample mean direction. That is, θ = T1 .
The maximum likelihood of κ is the solution to
A1 (κ ) = R
where
I1 ( x )
A1 ( x ) = .
I0 ( x )
That is, the MLE of κ is given by
κ * = A1−1 ( R )
This can be approximated by (see Fisher (1993) page 88 and Mardia & Jupp (2000) pages 85-86)
 5R 5
 2 R + R 3
+ R < 0.53
 6
0.43
κ * = − 0.4 + 139. R+ 0.53 ≤ R < 0.53
 1− R
 1
R ≥ 0.85
 3R − 4 R 2 + R 3
This estimate is very biased. This bias is corrected by using the following modified estimator.
  * 2 
maxκ − , 0 κ * < 2

 nκ *

n ≤ 15
κ =  ( n − 1) κ
3 *
κ ≥2
*


n n 2
(
+ 1 )
 κ* n > 15

Test for a Specified Mean Direction of Von Mises Data

There are several different hypothesis tests that have been proposed for testing H 0 :θ = θ 0 versus H1:θ ≠ θ 0 ,
where θ 0 is a specific value of the mean direction. The tests presented here require the additional assumption that
the data follow the Von Mises distribution, at least approximately.
It will be useful to adopt the following notation.
1 n
C* = ∑ cos(ai − θ0 )
n i=1
1 n
S* = ∑ sin(ai − θ0 )
n i=1

R* = [ S ] + [C ]
* 2 * 2

230-6
© NCSS, LLC. All Rights Reserved.
NCSS Statistical Software NCSS.com
Circular Data Analysis

Score Test
The score test, given by Mardia & Jupp (2000) page 123, is computed as
nκ
( )
2
χ S2 = S*
A1 (κ )


For large n, χ S2 follows the chi-square distribution with one degree of freedom.

Likelihood Ratio Test

The likelihood ratio test, given by Mardia & Jupp (2000) page 122, is computed as



[( ) ( )]
 4n R * 2 − C * 2
if n ≥ 5 and C * ≤ 2 / 3
( )
2
 2− C*
χL = 
2

1 − C * ( ) 
2
 2n 3 
 2 log if n ≥ 5 and C * > 2 / 3
( ) ( )
* 2 1− R* 2
 n + nC + 3n  

The test statistic, χ L2 , follows a chi-square distribution with one degree of freedom.

Watson & Williams Test

The Watson and Williams test, given by Mardia & Jupp (2000) page 123, is computed as
R* − C *
F= if C * ≥ 5 / 6
( )
1 − R / ( n − 1)
*

The test statistic, F, follows an F distribution with one and n-1 degrees of freedom.
Stephens Test
This test, given by Fisher (1993) pages 93-94, is computed as
sin(T1 − θ0 )
E=
1 / ( nκR )

If κ ≥ 2 , E follows the standard normal distribution.

Confidence Interval for Mean Direction assuming Von Mises

A general confidence interval for θ was given above. When the data can be assumed to follow a von Mises
distribution, a more appropriate interval is given by Mardia & Jupp (2000) page 124 and Upton & Fingleton
(1989) page 269. This confidence interval is given by

T1 ± cos −1 [
 2n 2 R 2 − nz 2 
 α  ] if R ≤ 2 / 3
 R 4n − z 

2 2
α(  )
 2 
 n 2 − n 2 − R 2 exp zα  
( )
  n
T1 ± cos −1   if R > 2 / 3
 R 
 
 

230-7
© NCSS, LLC. All Rights Reserved.
NCSS Statistical Software NCSS.com
Circular Data Analysis

Test for a Specified Concentration of Von Mises Data

Suppose you want to test a one-sided hypothesis concerning κ , given that the data come from a Von Mises
distribution and that the mean direction parameter is unknown. Fisher (1993) page 95 suggests the following
procedure when κ ≥ 2 .
When testing κ = κ 0 versus κ < κ 0 , reject the null hypothesis if

χ n2−1;α  1 3 
R <1−  + 2
2n  κ 0 8κ 0 
When testing κ = κ 0 versus κ > κ 0 , reject the null hypothesis if

χ n2−1;1−α  1 3 
R >1−  + 2
2n  κ 0 8κ 0 
These tests are based on the result that
2n(1 − R )
~ χ n2−1
1 3
+
κ 0 8κ 02

Confidence Interval for Concentration of Von Mises

An approximate confidence interval for κ when κ > 2 was given by Mardia & Jupp (2000) pages 126-127 as
 1 + 1 + 3b 1 + 1 + 3d 
 , 
 4b 4d 
where
n(1 − R )
b=
χ n2−1,1−α / 2

n(1 − R )
d=
χ n2−1,α / 2

Goodness of Fit Tests for the Von Mises Distribution

Stephens Modified Watson’s Test
The following goodness-of-fit test, published by Lockhart & Stephens (1985) as a modification of the Watson test
for the circle, is outlined in Fisher (1993) page 84. The test is conducted by calculating U 2 and comparing it to a
table of values. If the calculated value is greater than the critical value, the null hypothesis of Von Misesness is
rejected. Note that the test is only valid for samples of at least 20 angles.
The calculation of U 2 is as follows
2
2i − 1 
2
   1
n
1
U = ∑  p ( i ) −
2
− n  p −  +
i =1  2n   2  12n

230-8
© NCSS, LLC. All Rights Reserved.
NCSS Statistical Software NCSS.com
Circular Data Analysis

where
n

∑ p (i)
p = i =1
n

(
p ( i ) = Fκ a( i ) − T1 )
a(1) ≤ a( 2 ) ≤ a( 3) ≤ ≤ a( n ) are the sorted angles and Fκ (a − θ ) is the cumulative distribution function of the von
Mises distribution. Note that maximum likelihood estimates of κ and θ are used in the distribution function.
Lockhart & Stephens (1985) present a table of critical values that has been entered into NCSS. When a value of
U 2 is calculated, the table is interpolated to determine its significance level.
Cox Test
Mardia & Jupp (2000) pages 142-143 present a von Mises goodness-of-fit test that was originally given by Cox
(1975).
The test statistic, C, is distributed as a chi-squared variable with two degrees of freedom under the null hypothesis
that the data follow the von Mises distribution. It is calculated as follows.

sc2 ss2
C= +
nv c (κ ) nv s (κ )
where
n
sc = ∑ cos 2(a i − T1 ) − nα 2 (κ )
i =1

n
ss = ∑ sin 2(a i − T1 )
i =1

vc ( x ) =
1 + α4 [α / 2 + α 3 / 2 − α1α 2 ]
− α 22 − 1
2

2 (1 + α 2 ) / 2 − α12

vS ( x) =
α1 − α 4
−
(α1 − α 3 )
2

2 1 − α2

Multi-Group Tests
Three multi-group tests are available for testing hypotheses about two or more groups. The nonparametric
uniform-scores test tests whether the distributions of the groups are identical. The Watson-Williams F test tests
whether a set of mean directions are equal given that the concentrations are unknown, but equal, given that the
groups each follow a von Mises distribution. The concentration homogeneity test tests whether the concentration
parameters are equal, given that the groups each follow a von Mises distribution.

230-9
© NCSS, LLC. All Rights Reserved.
NCSS Statistical Software NCSS.com
Circular Data Analysis

Mardia-Watson-Wheeler Uniform-Scores Test

Suppose you have g populations following any common distribution from which random samples are taken and
you wish to test whether these distributions are equal. Fisher (1993) page 122 and Mardia & Jupp (2000) pages
156-157 present a nonparametric test that is calculated as follows

Wg = 2∑
g
(C 2
Ri + S Ri2 )
i =1 ni

( ) ( )
ni ni g
where C Ri = ∑ cos γ ij , S Ri = ∑ sin γ ij , n = ∑ ni , and γ ij are the circular ranks of the corresponding
j =1 j =1 i =1

angles. The circular ranks are calculated using

2πrij
γ ij =
n
where the rij are the ranks of the corresponding aij .

If all ni are greater than 10, the distribution of Wg is approximately distributed as a chi-square with 2g-2 degrees
of freedom.
Since ranks are used in this test, ties become an issue. We have adopted the strategy of applying average ranks.
Note that little has been done to test the adoption of this strategy within the realm of circular statistics.

Watson-Williams High Concentration F Test

Suppose you have g Von Mises populations from which random samples are taken and you wish to test whether
their mean directions are equal. That is, you wish to test the null hypothesis
H 0 : θ1 = θ2 =... = θ g
Mardia & Jupp (2000) pages 134-135 present the Watson-William High-Concentration F Test that is calculated as
follows

 g 
 ∑ R j − R / ( g − 1)
 3   j =1 
FWW = 1 + 
 8κ   g

 n − ∑ R j  / (n − g)
 j =1 

where κ is the maximum likelihood estimate of the concentration based on R and

nj nj g g
Rj = C + S , Cj =
2
j
2
j ∑ cos(a ) , S i j = ∑ sin(a ) , i R = C +S , C =
2 2
∑C
j =1
j , S = ∑S
j =1
j , and
i =1 i =1
g
n = ∑n
j =1
j .

The distribution of FWW is approximately distributed as an F with g-1 and n-1 degrees of freedom when the
assumptions that κ 1 = κ 2 =... = κ g and that the distributions are Von Mises are made. The approximation also
requires that κ ≥ 1 .

230-10
© NCSS, LLC. All Rights Reserved.
NCSS Statistical Software NCSS.com
Circular Data Analysis

Multi-Group Concentration Homogeneity Test

Suppose you have g groups from which random samples are taken and you wish to test whether the
concentrations are equal. That is, you wish to test the null hypothesis
H 0 :κ1 = κ 2 =... = κ g
Mardia & Jupp (2000) page 139 presents such a test. It is divided into three cases.

Case I. R < 0.45

U1 is approximately distributed as a chi-square with g-1 degrees of freedom
2
 g 
g
∑ w j f j 
U 1 = ∑ w j f j2 −  g 
j =1

j =1
∑ wj
j =1

(
4 nj − 4 ) and
where w j =
3
(
f j = sin −1 2 R j 3 / 8 )
Case II. 0.45 ≤ R ≤ 0.70
U2 is approximately distributed as a chi-square with g-1 degrees of freedom
2
 g 
g
 ∑ w j h j 
U 2 = ∑ wjhj −
2  j =1 
g
j =1
∑ wj j =1

nj − 3 −1
. 
 R j − 1089
where w j = and h j = sinh  
0.797449  .258 

Case III. R > 0.70

U3 is approximately distributed as a chi-square with g-1 degrees of freedom
  g
 
 n − ∑ Rj  g
1 

   n j − R j  
 ∑
j =1
U3= ν log − ν j log  
1+ d  ν j =1  νj 
 
   

1  g 1 1 
where ν j = n j − 1 , ν = n − g ,and d = ∑ −  .
3( g − 1)  j =1 ν j ν 

230-11
© NCSS, LLC. All Rights Reserved.
NCSS Statistical Software NCSS.com
Circular Data Analysis

Circular Correlation Measure

This section discusses a measure of the correlation between two circular variables presented by Jammalamadaka
and SenGupta (2001). Suppose a sample of n pairs of angles (a11 , a21 ),(a12 , a22 ),...,(a1n , a2 n ) is available. The
circular correlation coefficient is calculated as
n

∑ sin(a 1k − T1,1 ) sin(a2 k − T2,1 )

rc = n
k =1
n

∑ sin (a
k =1
2
1k − T1,1 )∑ sin 2 (a2 k − T2,1 )
k =1

where T1,1 is the mean direction of the first circular variable and T2,1 is the mean direction of second.

The significance of this correlation coefficient can be test using the fact the zr is approximately distributed as a
standard normal, where

nλ20λ02
zr = rc
λ22
and
1 n
λij = ∑ sin i (a1k − T1,1 ) sin j (a2 k − T2,1 )
n k =1

Data Structure
The data consist of one or more variables. Each variable contains a set of angular values. The rows may be
separated into groups using the unique values of an optional grouping variable. An example of a dataset
containing circular data is Circular1.S0. Missing values are entered as blanks (empty cells).

230-12
© NCSS, LLC. All Rights Reserved.
NCSS Statistical Software NCSS.com
Circular Data Analysis

Rose Plot / Circular Histogram Window Options

This section describes the specific options available on the Rose Plot / Circular Histogram Format window, which
is displayed when a Rose Plot / Circular Histogram Format button is clicked. Common options, such as axes,
labels, legends, and titles are documented in the Graphics Components chapter.

Rose Plot Tab

Data Type
The data type of the plot is specified independently of the data type specified on the Variables tab of the Circular
Data Analysis procedure.

230-13
© NCSS, LLC. All Rights Reserved.
NCSS Statistical Software NCSS.com
Circular Data Analysis

Direction
This option indicates whether the orientation of the plot is in a 'Clockwise' or 'Counter-Clockwise' direction.

Reference Angle (Rotation)

This option lets you indicate the position of 0 degrees by entering an offset angle. On the default circle, 0 degrees
is on the right (east), 90 degrees is at the top (north), 180 degrees is on the left (west), and 270 degrees is at the
bottom (south). This option lets you add an 'offset' to each angle which moves the position of 0 degrees around
the circle.
The offset must be between 0 and 360 degrees.

230-14
© NCSS, LLC. All Rights Reserved.
NCSS Statistical Software NCSS.com
Circular Data Analysis

Interior Objects
The two choices for plot styles are Rose Plot and Circular Histogram.

Group Display
When the data is grouped data, this option determines whether the petals within a bin are side-by-side, stacked
upon each other, or overlaid.
Side-by-Side
The bin width is divided equally by the number of groups and the petals are laid out sequentially in the bin.
Although the petals are narrower, they still encompass the points of the group that within the boundaries of the
whole bin.
Stacked
A single petal in each bin is divided by the number of groups. Rose plots with the group display set to Stacked
may be misleading because the proportional area is larger for the outside groups.
Overlaid
Each petal for each group is overlaid in each bin. Some degree of transparency is recommended when using the
Overlaid group display. It is also difficult to distinguish groups when there are more than 2 or 3 groups.

Radius for Interior Objects

This option specifies the distance to the outer edge of the bins and petals of the rose plot or circular histogram.

Petal Width
Specify the percent of the total width of each bin that is to be used for each petal.

Percent for Histogram Base

This is the percent of Radius for Interior Objects that is used for the base of the circular histogram.

230-15
© NCSS, LLC. All Rights Reserved.
NCSS Statistical Software NCSS.com
Circular Data Analysis

Number of Bins
Specify the number of bins for the circle.

First Bin Start Angle

This permits the user to change the angle at which the binning begins. This is useful, for example when the Data
Type is set to Compass, since this option can be used to center the bins on the directions. When Data Type is set
to Compass, the recommended First Bin Start Angle is 45 for 4 bins, 22.5 for 8 bins, and 11.25 for 16 bins.

Number of Radial Axes

Select the number of axes that go from the center to the outer edge of the interior region. If this is set to the same
number as the number of bins, these axes show the edges of the bins. The radial axes also begin at the First Bin
Start Angle.

Alternate Bin Fill

Check this box to show a background fill for each bin. The fills alternate beginning with Fill 1. When the number
of bins is odd, the adjacent first and last bins will both have Fill 1.

230-16
© NCSS, LLC. All Rights Reserved.
NCSS Statistical Software NCSS.com
Circular Data Analysis

Data & Means Tab

Raw Data Symbols

Check this box to show the raw data symbols.

Number of Raw Data Symbol Bins

Specify the number of bins for the raw data points. To use no binning set this to 0 or All Uniques.

First Bin Start Angle

This permits the user to change the angle at which the binning begins. This is useful, for example when the Data
Type is set to Compass, since this option can be used to center the bins on the directions.
When Data Type is set to Compass, the recommended First Bin Start Angle is 45 for 4 bins, 22.5 for 8 bins, and
11.25 for 16 bins.

Radius for Raw Data Symbols

This is the distance from the center at which the symbols are shown.

Width for Multiple Symbols at One Location

This specifies the width of the band that contains the symbols when there is more than one value at some
locations.

230-17
© NCSS, LLC. All Rights Reserved.
NCSS Statistical Software NCSS.com
Circular Data Analysis

Mean Symbols and Lines

Use these options to set up the visual representation of the circular means for each group.

References Tab

Direction References
The options in this section allow you to specify the tick marks and references going around the plot.

Magnitude References
The options in this section allow you to specify the tick marks and references going from the center to the outside
of the plot.

230-18
© NCSS, LLC. All Rights Reserved.
NCSS Statistical Software NCSS.com
Circular Data Analysis

Example 1 – Analysis of Circular Data

This section presents an example of how to run this procedure. The data are wind directions of two groups. The
data are found in the Circular1 dataset.

Setup
To run this example, complete the following steps:

1 Open the Circular1 example dataset

• From the File menu of the NCSS Data window, select Open Example Data.
• Select Circular1 and click OK.

2 Specify the Circular Data Analysis procedure options

• Find and open the Circular Data Analysis procedure using the menus or the Procedure Navigator.
• The settings for this example are listed below and are stored in the Example 1a and Example 1b settings
templates. To load one of these templates, click Open Example Template in the Help Center or File
menu.

Option Value
Variables Tab
Data Variables ........................................ Wind
Grouping Variable ................................... Group
Hypothesized Theta ................................ 40
Hypothesized Kappa .............................. 2

3 Run the procedure

• Click the Run button to perform the calculations and generate the output.

Summary Statistics Section

Summary Statistics Section - Wind ────────────────────────────────────────────────

Mean Circular
Sample Mean Resultant Circular Standard Circular Von Mises
Size Direction Length Variance Deviation Dispersion Concentration
Group (N) (Theta) (R bar) (V) (v) (Delta) (Kappa)
1 10 41.5869 0.9324 0.0676 21.4299 0.1449 5.5452
2 10 42.6725 0.9599 0.0401 16.3991 0.0768 9.1850

Notes:
Theta is the mean direction.
R bar is a measure of data concentration. 0 <= R bar <= 1. R bar = 1 implies high concentration.
V = 1 - R bar, so 0 <= V <= 1. V = 0 implies high concentration.
v is the circular analog of the linear standard deviation.
Delta is another measure of circular spread.
Kappa is the concentration parameter of the Von Mises distribution.

Group
This is the group (or variable) presented on this line.
Sample Size
This is the number of nonmissing values in this group.

230-19
© NCSS, LLC. All Rights Reserved.
NCSS Statistical Software NCSS.com
Circular Data Analysis

Mean Direction
This is estimated mean direction, T1 .

Mean Resultant Length

This is the estimated mean resultant length, R1 . It is a measure of data concentration. An R1 close to zero implies
low data concentration. An R1 close to one implies high data concentration.

Circular Variance
The circular variance, V, is a measure of variation in the data. Note that V = 1 − R1 .

Circular Standard Deviation

The circular standard deviation is v = − 2 ln( R1 ) . Note that it is not the square root of the circular variance.

Circular Dispersion
1 − T2
The circular dispersion, δ = , is another measure of variation.
2 R12

Von Mises Concentration

This is the estimated concentration parameter of the von Mises distribution, κ .

Mean Direction Section

Mean Direction Section - Wind ───────────────────────────────────────────────────

Lower 95.0% Upper 95.0% Standard

Sample Mean Confidence Confidence Error of
Size Direction Limit Limit Mean
Group (N) (Theta) of Theta of Theta Direction
1 10 41.5869 27.9417 55.2321 6.8964
2 10 42.6725 32.7516 52.5934 5.0365

Notes:
This large-sample confidence interval does not require a von Mises distribution assumption.

This report provides the large sample confidence interval for the mean direction as described by Upton &
Fingleton (1989) page 220. Note that this interval does not require the assumption that the data come from the von
Mises distribution.

230-20
© NCSS, LLC. All Rights Reserved.
NCSS Statistical Software NCSS.com
Circular Data Analysis

Variation Statistics Section

Variation Statistics Section - Wind ────────────────────────────────────────────────

Circular
Sample Circular Standard Circular
Size Variance Deviation Dispersion Skewness Kurtosis
Group (N) (V) (v) (Delta) (s) (k)
1 10 0.0676 21.4299 0.1449 -0.0795 -1.7244
2 10 0.0401 16.3991 0.0768 -4.8582 5.4248

Notes:
These statistics are designed to assess and compare the variation in the data.
V = 1 - R bar, so 0 <= V <= 1. V = 0 implies high concentration.
v is the circular analog of the linear standard deviation.
Delta is another measure of circular spread.
For symmetric, unimodal data sets, the skewness is close to zero.
For von Mises data sets, the kurtosis is close to zero.

This report provides measures of data variation and dispersion which were defined in the Statistical Summary
Report. It also provides measures of the skewness and kurtosis of the data.
Skewness
This is a measure of the skewness (lack of symmetry about the mean) in the data. Symmetric, unimodal datasets
have a skewness value near zero.
Kurtosis
This is a measure of the kurtosis (peakedness) in the data. Von Mises datasets have a kurtosis near zero.

Von Mises Distribution Estimation Section

Von Mises Distribution Estimation Section - Wind ──────────────────────────────────────

Lower 95.0% Upper 95.0% Lower 95.0% Upper 95.0%

Sample Mean Confidence Confidence Von Mises Confidence Confidence
Size Direction Limit Limit Conc. Limit Limit
Group (N) (Theta) of Theta of Theta (Kappa) of Kappa of Kappa
1 10 41.5869 29.1135 54.0603 5.5452 2.3222 8.4051
2 10 42.6725 33.2995 52.0455 9.1850 3.0227 13.2191

Notes:
These large-sample confidence intervals assume that the data follow a von Mises assumption.
The confidence interval for kappa requires the estimated kappa to be > 2.

This report provides estimates and confidence intervals of the parameters (mean direction and concentration) of
the von Mises distribution that best fits the data. Note that the von Mises distribution is a symmetric, unimodal
distribution. You should check the rose plot or circular histogram to determine if the data are symmetric.
The formulas used in the estimation and confidence intervals were given earlier in this chapter. They come from
Mardia & Jupp (2000).

230-21
© NCSS, LLC. All Rights Reserved.
NCSS Statistical Software NCSS.com
Circular Data Analysis

Trigonometric Moments Section

Trigonometric Moments Section - Wind ─────────────────────────────────────────────

Mean Mean Mean Mean

Group N Cos(a) Sin(a) Cos(2a) Sin(2a) R bar 2R bar Theta 2Theta
1 10 0.6974 0.6189 0.0903 0.7426 0.9324 0.7481 41.5869 83.0670
2 10 0.7057 0.6506 0.1085 0.8516 0.9599 0.8585 42.6725 82.7373

Notes:
These values are used in the calculation of other statistics.

This report provides summary statistics that are used in other calculations.
Mean Cos(a)
1 n
This is C1 = ∑ cos(ai ) .
n i =1
Mean Sin(a)
1 n
This is S1 = ∑ sin(ai ) .
n i =1
Mean Cos(2a)
1 n
This is C2 = ∑ cos(2ai ) .
n i =1
Mean Sin(2a)
1 n
This is S 2 = ∑ sin(2ai ) .
n i =1
R bar

This is R1 =
1
n
(
n C12 + S12 . )
2R bar

This is R2 =
1
n
(
n C22 + S22 . )

230-22
© NCSS, LLC. All Rights Reserved.
NCSS Statistical Software NCSS.com
Circular Data Analysis

Theta, 2 Theta
This is calculated using the following formula with p set to 1 and then 2, respectively.

 −1  S p 
tan   C p > 0, S p > 0
  Cp 
  Sp 

Tp = tan −1   + π Cp < 0
  Cp 
 S 
tan −1  p  + 2π S p < 0, C p > 0
  Cp 

Multiple-Group Hypothesis Tests Section

Multiple-Group Hypothesis Tests Section - Wind ───────────────────────────────────────

Null
Hypothesis Test Test Prob Reject H0
(H0) Name Statistic Level at 0.05 Level
Equal Distributions Uniform Scores Test 6.7392 0.0344 Yes
Equal Directions Watson-Williams F Test 0.0147 0.9047 No
Equal Concentrations Concentration Homogeneity Test 0.5717 0.4496 No

Notes:
These statistics test various hypotheses about the parameters of von Mises distributions.
They require that each group follow the von Mises distribution.
The Uniform Scores test requires samples of at least 10.
The Watson-Williams F-test assumes that all kappa's are equal and that their average is > 1.

This report provides tests for three hypotheses about the features of several von Mises datasets. That is, it
provides a test of whether the distributions are identical, whether the mean directions are identical, and whether
the concentrations are identical. These tests are documented in the Technical Details section of this chapter.

Two-Group Hypothesis Tests Section

Two-Group Hypothesis Tests Section - Wind ─────────────────────────────────────────

─ Equal Distributions ─ ─ Equal Directions ─ ─ Equal Concentrations ─

First Second Test Prob Test Prob Test Prob
Group Group Statistic Level Statistic Level Statistic Level
1 2 6.7392 0.0344 0.0147 0.9047 0.5717 0.4496

Notes:
These statistics test various hypotheses about the parameters of von Mises distributions.
They require that each group follow the von Mises distribution.
Equal distributions tested by the Mardia-Watson-Wheeler uniform scores test. Requires all Ni > 10.
Equal directions tested by the Watson-Williams F test. Assumes Von Mises data with equal kappa's, all > 1.
Equal concentrations tested by concentration homogeneity test. Assumes Von Mises data.

This report provides the same three tests as the Multiple-Group Hypothesis Tests Section, taken two groups at a
time. It allows you to pinpoint where differences occur.

230-23
© NCSS, LLC. All Rights Reserved.
NCSS Statistical Software NCSS.com
Circular Data Analysis

Tests for a Specified Mean Direction Assuming Von Mises Data – Test
Statistic & Prob Levels
Tests for a Specified Mean Direction Assuming Von Mises Data - Test Statistics - Wind ─────────────

Actual H0 Score Likelihood Watson & Stephens

Sample Mean Mean Test Ratio Williams Test
Size Direction Direction Z CS F Z
Group (N) (Theta) (Theta0) Value Value Value Value
1 10 41.5869 40.0000 0.0409 0.0470 0.0476 0.0397
2 10 42.6725 40.0000 0.1949 0.2266 0.2341 0.1917

Notes:
These procedures test whether the mean direction is equal to a specified value, when kappa (concentration) is
unknown.
They assume that the data follow the von Mises distribution.
The Score Test requires a large sample size.
The Likelihood Ratio Test requires a sample size of at least 5.
The Watson & Williams Test requires a large value of kappa.
The Stephens Test requires kappa to be greater than 2.

Tests for a Specified Mean Direction Assuming Von Mises Data - Probability Levels - Wind ───────────

Actual H0 Score Likelihood Watson & Stephens

Sample Mean Mean Test Ratio Williams Test
Size Direction Direction Prob Prob Prob Prob
Group (N) (Theta) (Theta0) Level Level Level Level
1 10 41.5869 40.0000 0.8398 0.8284 0.8321 0.8422
2 10 42.6725 40.0000 0.6589 0.6340 0.6400 0.6615

Notes:
This report gives the probability levels of the test statistics displayed in the previous report.
Although the probability levels of four tests are given, you should use only one of these.

This section reports the results of four tests of the hypothesis that the mean direction of a particular group is equal
to a specific value. These are two-sided tests. They were documented earlier in this chapter.
The first table gives the values of the test statistics. The second table gives the probability levels. The null
hypothesis is rejected when the probability level is less than 0.05 (or some other appropriate cutoff).

Tests for a Specified Concentration Assuming Von Mises Data

Tests for a Specified Concentration Assuming Von Mises Data - Wind ─────────────────────────

Prob Prob
Sample Actual H0 Level of Level of
Size Concentration Concentration Chi-Square (H1:Kappa (H1:Kappa
Group (N) (Kappa) (Kappa0) Value < Kappa0) > Kappa0)
1 10 5.5452 2.0000 2.2756 0.0137 0.9863
2 10 9.1850 2.0000 1.3518 0.0019 0.9981

Notes:
These statistics test whether the kappa (concentration) parameter is equal to the specified value.
The tests require that the estimated kappa is > 2.

This section reports the results of two, one-sided tests of the hypothesis that the concentration parameter of each
group is equal to a specific value. They were documented earlier in this chapter.
The first probability level is for testing the null hypothesis that kappa is greater than or equal to kappa0. The
second probability level is for test the null hypothesis that kappa is less than or equal to kappa0.

230-24
© NCSS, LLC. All Rights Reserved.
NCSS Statistical Software NCSS.com
Circular Data Analysis

Uniform Distribution Goodness-of-Fit Tests

Uniform Distribution Goodness-of-Fit Tests - Wind ─────────────────────────────────────

Rayleigh's Rayleigh's Kuiper's Kuiper's Watson's Watson's

Sample Test Test Test Test Test Test
Size Statistic Prob Statistic Prob Statistic Prob
Group (N) (S*) Level (V) Level (U2) Level
1 10 20.2993 0.0000 2.7145 0.0000 0.5657 0.0001
2 10 21.7499 0.0000 2.8088 0.0000 0.6788 0.0000

Notes:
The tests in this report assess the goodness-of-fit of the uniform distribution.
The Rayleigh test requires samples of at least 20.
The Kuiper and Watson tests require samples of at least 8.

This section reports the results of three goodness-of-fit tests for the uniform distribution. They were documented
earlier in this chapter.
These tests may be viewed as testing whether the data are distributed uniformly around the circle.

Von Mises Distribution Goodness-of-Fit Tests

Von Mises Distribution Goodness-of-Fit Tests - Wind ────────────────────────────────────

Watson's Watson's Cox's Cox's

Sample Test Test Test Test
Size Statistic Prob Statistic Prob
Group (N) (U2) Level (S) Level
1 10 0.0340 0.5000 0.4030 0.8175
2 10 0.1282 0.0322 2.9309 0.2310

Notes:
The tests in this report assess the goodness-of-fit of the von Mises distribution.
Both tests require samples of at least 20.

This section reports the results of two goodness-of-fit tests for the von Mises distribution. They were documented
earlier in this chapter. Several hypothesis tests assume that the data follow a von Mises distribution. These tests
allow you to check the accuracy of this assumption.

230-25
© NCSS, LLC. All Rights Reserved.
NCSS Statistical Software NCSS.com
Circular Data Analysis

Rose Plots
Combined Rose Plot ─────────────────────────────────────────────────────────

Individual Rose Plots ─────────────────────────────────────────────────────────

These plots show the distribution of the data around the circle.

230-26
© NCSS, LLC. All Rights Reserved.
NCSS Statistical Software NCSS.com
Circular Data Analysis

Circular Histograms
Combined Rose Plot ─────────────────────────────────────────────────────────

Individual Rose Plots ─────────────────────────────────────────────────────────

The circular histograms are generated by clicking the Plot Format buttons and setting Interior Objects to “Circular
Histogram.”

230-27
© NCSS, LLC. All Rights Reserved.