A C T A U N I V E R S I T A T I S L O D Z I E N S I S
FO L IA O EC O N O M IC A 141, 1997
Tadeusz Gerstenkorn* Joanna Gerstenkorn**
R E M A R K S ON T H E G E N E R A L IZ E D D O U B LY T R U N C A T E D G A M M A D IST R IB U T IO N
Abstract. In the present paper we discuss a doubly truncated generalized gam m a
distribution and give form ulae for the m om ents o f this distribution and special cases together with examples o f calculations.
Key words: truncated probability distributions, m om ents, generalized gam m a distribution.
1. IN T R O D U C T IO N
After introducing the generalized gam m a distribution to the literature by S t a c y (1962), the investigations of this distribution were carried out by m any authors such as Š r ó d k a (1964, 1966, 1967, 1970), S t a c y and M i h r a n (1965), M a l i k (1967), W a s i l e w s k i (1967), H a r t e r ( 1967), L i e n h a r d and M e y e r (1967), R o s ł o n e k (1968), H a g e r and B a i n (1970), P o d o l s k i (1972), K r ó l i k o w s k a (1973), J a k u s z e n k o w (1973, 1974, 1976), L a j k ó (1977), L a w l e s s (1980), A c h e a r and B o l - f a r i n e (1986), M a s w a d a h (1991).
The im portance o f this distribution lies in the fact it is a generalization o f the gamma, Weibull, Rayleigh and X distributions significant in m any technical and economic applications.
In the present paper we deal with a generalized doubly truncated gam m a distribution and give form ulae for the m om ents o f this distribution together with examples of calculations. This will enable us to apply the distribution and its special cases whenever the range o f the variable is essentially bounded to some interval o f values. In the practice of operation researches, these m ay be the cases quite frequently encountered.
* U niversity o f Łódź, Faculty o f M athem atics. ** U niversity o f Łódź, C hair o f Statistical M ethods.
2. T H E R ESU LTS
The density function of random variable X with the generalized gam m a distribution is given by
A part from m any articles in journals, distribution (1) is also easy to find in the lexicon M ü l l e r et al. 1975 and handbook G e r s t e n k o r n , Ś r o d k a (1983) listed in the references.
As some special cases of (1) we have: 1) gam m a distribution if a = 1, 2) Weibull distribution if a = p, 3) Rayleigh distribution if a = p — 2, 4) x distribution if a = 2, p = n,
(most frequently, a = 1/2, see M ü l l e r et al. 1975, p. 34),
a) a = 2, p = n = 2 - the Maxwell distribution with 2 com ponents (most frequently, X = 1/2a 2).
b) a = 2, p = n = 3 - the Maxwell distribution with 3 com ponents (most frequently, X = 1/2<т2).
We assume th at f ( x ) is a positive density function o f random variable
X with P ( X e ( c , í/)) = 1. The density function f ^ u ) o f random variable U,
given by
is called the density of a doubly truncated distribution of random variable
X. We consider a truncated distribution if the values of random variable
have to be limited to interval [a, b].
D istributions 1-4 are well know n and have been investigated in m any papers.
F o r distribution (1), we have from (2)
for jc > 0; a, p, X > 0, for a: < 0.
(
1)
ь 0 for и < a or и > b a(
2
)
ь f ď 1e A“7J x p 1e lx’dx o f 0 < a < и sg b, /i(m ) = S 0 o f и < a or u > b.D enote
h
W„ = f vp~ i+ne ~ iy'dv, и = 0, 1 , 2 , ... (4)
a
Lemma. F o r (4), we have
W„ = j - [ a p+n~ae ~ Álŕ - b p+n- ae - Xa* + (p + n - a)ITn_ J . (5)
Proof. We can write 1 b
W„= - - r f-A<xva- 1e - iv"v',+n- adv. Aoc a
A fter integration by parts, we have
1 b
W„ = - T [(vp+n- ae - Xv’) l ba- ( p + n - a ) f v p- 1+n- ae - Xv‘dv] = (5)
A C C a
We denote by m n the m om ent of order n o f random variable U with density function (3), th at is,
m n = E i l ß ) = J unJ\{u)du. (6)
a
From (6) and (4) we immediately have
Theorem. If random variable U has density function (3) o f the truncated generalized gam m a distribution (1), then, for и > 1,
m n = W„/W0 (7)
where W„ is given by (4). Examples.
We calculate m 2 in the ease of truncated generalized gam m a distribution (3). F rom (7) and (5) we have
m 2 = T [ар+2- ае ~ Ха’ - b p+2- * e - Xb‘ + (p + 2 - a) W2- e]. Aoc VYq
We consider the following special cases: 1. Let a = 1, Я = 1; then (gam m a distribution)
m 2 = (ap+íe~a - b p+1e~b + (p + 1 ) fV1) / W 0 =
W e notice that
W Q - J vp_ 1e ~ vdv = Г (p, b) - Г(/7, a) (8)
a
where Г (p, x ) represents the so-called incomplete gam m a function
Г(р, a) = J- vp~ i e~'dv,
о
which means that one can calculate (8) with the aid e.g. K . Pearson’s Tables. 2. If p = 1 = a (special case of the W eibull distribution), then
W 0 = e - a - e ~ b,
therefore
m 2 = {a2e ~ 2 — b2e~b + 2[(a + \)e~a - (b + 1 )е “ ь]}/(<Га - e " b) =
= {(a2 + 2a + 2)e~a - (b2 + 2b + 2) e - b}l(e~a - e~b). 3. Let a = p = 2 (Rayleigh distribution). In this case
m 2 = W 2I W 0 = 2^ p r ( a 2e _A“J - b2e ~Xb’ + 2 ^ 0),
however,
Ж
0 =
}ve~l *dv= - ^
“AvJ)la =
therefore
(a2 + ^ ) e - Aal- ( f c 2 + ^ ) e - A'’1
ffl2 = g-й * _ g-Abä •
4. Let а = 2, p = 3 (Maxwell distribution with 3 com ponents, then
т г= ~ b 3 e ~ Xb' + 3 W o ) where, in this case
W 0 = J v2e ~ Xv'dv.
a
A fter integration by parts, we have
and, further,
ь I JlXb [Z __ __
J e ~ Xv'dv = J е" ,2/2Л = / - [Ф(72Я6) - Ф(^2Яа)]. я \/2Я ,/2Аа V ^
Using the tables o f the norm al distribution Ф(/), we obtain the required result.
R E FE R E N C E S
A c h c a r J. A. and B o l f a r i n e H. (1986): The log-linear m odel with a generalized gamma
distribution fo r the error: a Bayesian approach, „Statistics and P roba bility L etters” , 4, 6,
p. 325-332.
G e r s t e n k o r n T., Ś r o d k a T. (1983): Combinatorics and probability theory, Polish Scientific Publishers, 7th ed. W arsaw, (in Polish).
H a g e r H. W. and B a i n L. J. (1970): Inferential procedures fo r the generalized gamma
distribution, JASA 65, p. 1601-1609.
H a r t e r H . L. (1967): M axim um likelihood estimation o f the parameters o f a four-parameter
generalized gamma population fo r complete and censored samples, „Technom etrics” 9,
p. 159-165.
J a k u s z e n k o w H. (1976): Bayes estimator in a generalized gamma distribution, Z eszyły N aukow e, Politechnika Łódź, (Sei. Bull. Łódź Techn. Univ.), 245, M al. 8, p. 25-33. J a k u s z e n k o w H. (1974): On properties o f the generalized gamma distribution, D em onstratio
M ath., 7, 1, p. 13-22.
J a k u s z e n k o w H . (1973): On some property o f the generalized gamma distribution, Ann. Soc. M ath. Polon., Ser. I, Comm ent. M ath. Prace M at., 17, 1, p. 237-242.
K r ó l i k o w s k a К . (1973): On the characterization o f some fam ilies o f distributions, Ann. Soc. M ath. Polon. Ser. I, Comm ent. M ath. Prace M at., 17, 1, p. 243-261.
L a j k ó К . (1977): A characterization o f generalized normal and gamma distributions, Colloq. M ath. Soc. János Bolyai, 21. Analytic F unction M ethods in Probability Theory, Debrecen (Hungary).
L a w l e s s J. F. (1980): Inference in the generalized gamma and log-gamma distributions, „Technom etrics” 22, p . 409-419.
L i e n h a r d J . H. and M e y e r P. L. (1967): A physical basis fo r the generalized gamma
distribution, Q uart. A ppl. M ath ., 25, p. 330-334.
M a l i k H. J. (1967): E xact distribution o f the quotient o f independent generalized gamma
variables, Canad. M ath. Bull., 10, 3, p. 463-465.
M a s w a d a h M . S. (1991): Structural inference on the generalized gamma distribution based
on type II progressively censored sample, J.A ustral. M ath. Soc., Ser. A, 50, 1, p. 15-22.
M ü l l e r P. H. und Autorenkollektiv (1975): Lexikon der Stochastik, Wahrscheinlichkeitsrechnung
und Mathematische Statistik, Akademie Verlag, 2. bearbeitete und erweiterte Auflage, Berlin.
P o d o l s k i H. (1972): The distribution o f a product o f n independent random variables with
generalized gamma distribution, D em onstratio M ath., 4, 2, p. 119-123.
R o s ł o n e k E. (1968): On some characterization o f the generalized gamma distribution, Zeszyty N auk. Politech. W arszaw., 173, p. 127-134.
S t a c y E. W. (1962): A generalization o f the gamma distribution, A nn. M ath. Stat., 28 p. 1187-1192.
S t a c y E. W. and M i h r a n G . A. (1965): Parameter estimation fo r a generalized gamma
distribution, „Technom etrics” 7, p. 349-358.
Ś r o d k a T. (1966): Composition o f the Laplace distribution with a generalized gamma, M axw ell
and Weibull distributions (in Polish), „Zeszyty N auk. Politech. Ł ódź” , 77, W łókiennictwo
14, p. 21-28.
Ś r o d k a T. (1970): Distribution o f the product o f powers o f two independent random variables
subject to generalized gamma distribution (in Polish), „Zeszyty N auk. Politech. Ł ódź” 112,
E lektryka 31, p. 47-66.
Ś r o d k a T. (1964): Estimateurs et intervalles de confiance d'un certain param etre dans la
distribution gamma généralisée, Rev. Statist. A ppl., 12, 2, p. 79-83.
Ś r o d k a T. (1967): On the distribution o f product and ratio o f powers o f two independent
random variables with the generalized gamma M axw ell and Weibull distribution, A nn. Soc.
M ath. Polon., Ser. I, Comment. M ath. Prace M at., 11, p. 77—85.
W a s i l e w s k i M . J. (1967): Sur certaines propriétés de la distribution gamma généralisée, Rev. Statist. A ppl., 15, 1, p. 95-105.
Tadeusz Gerstenkorn, Joanna Gerstenkorn
U W AG I O U O G Ó L N IO NY M PO D W Ó JN IE U CIĘTY M R O Z K Ł A D Z IE G A M M A
W pracy przedstaw iony jest podwójnie ucięty uogólniony rozkład gam m a i podane są wzory na m om enty tego rozkładu oraz jego przypadków szczególnych w raz z przykładam i ich wyliczeń.