1

PPLOT

PPLOT produces probability plots of one or more sequence or time series

variables. The variables can be standardized, differenced, and/or transformed
before plotting. Expected normal values or deviations from expected normal values
can be plotted. PPLOT can be used to investigate whether the data are from a
specified distribution: normal, lognormal, logistic, exponential, Weibull, gamma,
beta, uniform, Pareto, Laplace, half normal, chi-square and Student’s t.

Notation
The following notation is used throughout this chapter unless otherwise stated:

X Sample mean
S Sample standard deviation
LX 1 6
Sample mean for ln x i
LS 1 6
Sample standard deviation for ln x i
xi Value of the ith observation
x (i ) The ith smallest observation
Ri Corresponding rank for xi
n Sample size
frdist ( x i ) Fractional rank of xi for the specified distribution function
adist ( xi ) Score for the specified distribution function
α Location parameter
β Scale parameter
γ Shape parameter
ν Degrees of freedom

1
This procedure was introduced in SPSS 7.0 and replaces the NPPLOT procedure
of earlier releases.

1
2 PPLOT

Fractional Ranks
Based on the rank Ri for the observation xi , the fractional rank frdist ( x i ) is
computed and used to estimate the expected cumulative distribution function of X.
One of four methods can be selected to calculate the fractional rank frdist ( x i ) :

%K( R − 3 ) / (n + 1 ) Blom
KK( R − 81 ) / n 4
i

(x ) = & i Rankit
KK( R − 12 ) / (n + 1 )
frdist i
Tukey
K' R / (n3+ 1) 3
i

i Van der Waerden

Scores
The score of the specified distribution for case i is defined as

adist ( xi ) = Fdist −1 ( frdist ( xi )) i = 1, K, n

where Fdist −1 is the inverse cumulative specified distribution function.

P-P Plot
For a P-P plot, the fractional rank and the cumulative specified distribution function
Fdist are plotted:

2 fr 1 x 6, F 1 x 67
dist i dist i i = 1,K, n
 PPLOT 3

Q-Q Plot
For a Q-Q plot, the observations and the score for the specified distribution
function are plotted.

1x , a
i dist ( x i ) 6 i = 1,K, n

Distributions
The distributions and their parameters are listed below. Parameters may be either
specified by users or estimated from the data. Any parameter values specified by
the user should satisfy the conditions indicated.

Beta( β 1 , β 2 ) 0 5 0 5
β 1 > 0 and β 2 > 0 are scale parameters.

Chi-square(ν ) ν 0> 05 is the degrees of freedom.

Exponential( β ) β 0> 05 is a scale parameter.

Gamma( γ , β ) γ 0> 05 is a shape parameter and β 0> 05 is the scale

parameter.

Half Normal( β ) 0 5
β > 0 is a scale parameter and the location parameter
is 0.

Laplace(α , β ) 0 5
α is the location parameter and β > 0 is the scale
parameter.

Logistic(α , β ) 0 5
α is the location parameter and β > 0 is the scale
parameter.

Lognormal( β , γ ) 0 5 0 5
β > 0 is a scale parameter and γ > 0 is a shape
parameter.

Normal(α , β ) 0 5
α is the location parameter and β > 0 is the scale
parameter.

Pareto( β ,b); 0 5 0 5
β > 0 is scale parameter and b > 0 is an index of
inequality.
4 PPLOT

Student’s t(ν ) 0 5
v > 0 is the degrees of freedom specified by the user.

Uniform(a,b) a is a lower bound and b is an upper bound.

Weibull( β , γ ) 0 5 0 5
β > 0 is a scale parameter and γ > 0 is a shape
parameter.

Estimates of the Parameters

The estimates for parameters for each distribution are defined below.

Beta( β 1 , β 2 ) β$ 1 = X
%&
X (1 − X )
−1
()
' *
scale parameter
S2

%K X 31 − X 8 − 1(K
β$ 2 = (1 − X )&
K' S 2 )K*
scale parameter

Chi-square(ν ) ν is the degrees of freedom specified by the user.

β$ =
1
Exponential( β ) scale parameter
X

X2
Gamma( γ , β ) γ$ = 2 shape parameter
S

β$ = 2
X
scale parameter
S

Half Normal( β ) β$ = x12 +...+ x n2 scale parameter

Laplace(α , β ) α$ = X location parameter

S2
β$ = scale parameter
2

Logistic(α , β ) α$ = X location parameter

β$ = 3
S  
π  
, π = 3.1415927 scale parameter
 PPLOT 5

∧ −
Lognormal β = exp( L X ) scale parameter
∧
γ = LS shape parameter

Normal( α , β ) α$ = X location parameter

β$ = S scale parameter

Pareto( β ,b); ;
β$ = min x1 ,K, x n @ scale parameter
1
b$ = index of inequality
LX − ln β$ 49
Student’s t(ν ) v is the degrees of freedom specified by the user.

Uniform(a,b) ; @
a$ = min x1 ,K, x n lower bound

b$ = max ;x ,K, x @ 1 n upper bound

n
− −
∑ U Y − nU Y
i i
Weibull( β , γ ) β$ = i =1
n
scale parameter
−
∑ (U − U ) i
2

i =1
− ∧ − ∧
γ$ = exp ( −((Y − β U ) / β )) shape parameter

4 2
where Yi = ln − ln 1 − frdist x i 1 679 and U i 1 6
= ln xi

References
Kotz, S., and Johnson, N. L., eds. 1988. Encyclopedia of statistical sciences. John
Wiley & Sons, Inc.: New York.