4.6 The Gamma Probability Distribution
4.6 The Gamma Probability Distribution
and its expected value (mean), variance and standard deviation are,
p
µ = E(Y ) = αβ, σ 2 = V (Y ) = αβ 2 , σ = V (Y ).
One important special case of the gamma, is the continuous chi–square random vari-
able Y where α = ν2 and β = 2; in other words, with density
( ν−2 y
y 2 e− 2
2ν/2 Γ( ν2 )
, 0 ≤ y < ∞,
f (y) =
0, elsewhere,
and its expected value (mean), variance and standard deviation are,
p
µ = E(Y ) = ν, σ 2 = V (Y ) = 2ν, σ = V (Y ).
Another important special case of the gamma, is the continuous exponential random
variable Y where α = 1; in other words, with density
1 −y/β
β
e , 0 ≤ y < ∞,
f (y) =
0, elsewhere,
and its expected value (mean), variance and standard deviation are,
µ = E(Y ) = β, σ 2 = V (Y ) = β 2 , σ = β.
1. Gamma distribution.
R∞
i. Γ(1.2) = 0 y 1.2−1 e−y dy =
(choose one) (i) 0.92 (ii) 1.12 (iii) 2.34 (iv) 2.67.
PRGM GAMFUNC ENTER ENTER (again!) 1.2 ENTER
ii. Γ(2.2) ≈ (choose one) (i) 0.89 (ii) 1.11 (iii) 1.84 (iv) 2.27.
PRGM GAMFUNC ENTER ENTER (again!) 2.2 ENTER
iii. Notice
Γ(2.2) ≈ 1.11 = (2.2 − 1)Γ(2.2 − 1) = 1.2Γ(1.2) ≈ 1.2(0.92) ≈ 1.11.
In other words, in general,
Γ(n) = (n − 1)!
f(y) f(y)
0.2 (1) 0.2
(3) (4)
(2) (6)
(5)
y y
0 20 0 20
iii. If (α, β) = (2, 5), P (1.4 < Y < 2.7) = P (Y ≤ 2.7) − P (Y ≤ 1.4) ≈
(i) 0.07 (ii) 0.11 (iii) 0.21 (iv) 0.33.
Find PRGM GAMDSTR ENTER ENTER 2 ENTER 5 ENTER 2.7 ENTER
then subtract PRGM GAMDSTR ENTER ENTER 2 ENTER 5 ENTER 1.4 ENTER
(e) What is the 90th percentile waiting time; in other words, what is that time
such that 90% of waiting times are less than this time?
If P (Y < φ0.90 ) = 0.90, then φ0.90 ≈ (choose one)
(i) 1.89 (ii) 2.53 (iii) 3.72 (iv) 3.89.
PRGM GAMINV ENTER ENTER 2 ENTER 1 ENTER 0.9 ENTER
f(y) f(y)
2 4 6 8 10 2 4 6 8 10 y
y
3.6 7.0
3.9
(a) chi-square with 4 df (b) chi-square with 10 df
f(y) f(y)
0.72
0.72
y y
2 4 6 8 10 2 4 6 8 10
iv. The 32nd percentile is that waiting time such that 32% of the waiting
times are less than this waiting time and 68% are more than this time.
(i) True (ii) False
5. Exponential: waiting time for emails. Assume waiting times for emails follow
an exponential distribution,
1 −y/β
β
e , 0 ≤ y < ∞,
f (y) =
0, elsewhere,
122Chapter 4. Continuous Variables and Their Probability Distributions (ATTENDANCE 7)
(choose one) (i) e−10 (ii) e−20 (iii) e−30 (iv) e−40 .
9
Unlike the gamma and chi-square distributions, it is fairly easy to integrate exponential densities.
Section 7. The Beta Probability Distribution (ATTENDANCE 7) 123
(b) The chance batteries last at least 15 hours, given that they have already
lasted at least 5 hours is
P (Y > 15, Y > 5) P (Y > 15) 1 − (1 − e−3(15) )
P (Y > 15|Y > 5) = = = =
P (Y > 5) P (Y > 5) 1 − (1 − e−3(5) )
(choose one) (i) e−10 (ii) e−20 (iii) e−30 (iv) e−40 .
(c) In other words10 ,
and also
P (Y > s + t, Y > t) P (Y > s + t)
P (Y > s + t|Y > t) = =
P (Y > t) P (Y > t)
then, combining the last two equations,
P (Y > s + t)
= P (Y > s)
P (Y > t)
or
P (Y > s + t) = P (Y > t)P (Y > s)
− βa
But P (Y > a) = e , and so
s+t s t
e− β = e− β e− β
0 1 y 0 1 y
(a) increasing a, b = 3 (b) increasing a, b = 5
ii. If α and β become more equal, beta density becomes (choose one)
(i) more symmetric.
(ii) more skewed.
iii. If β decreases, “center” (mean) of beta density
(i) decreases.
(ii) remains the same.
(iii) increases.
Section 7. The Beta Probability Distribution (ATTENDANCE 7) 125
ii. For (α, β) = (2, 5), P (Y < 0.2) = F (0.2) ≈ (choose one)
(i) 0.34 (ii) 0.41 (iii) 0.67 (iv) 0.77.
PRGM BETADSTR ENTER ENTER 2 ENTER 5 ENTER 0.2 ENTER
iii. For (α, β) = (2, 5), P (0.1 ≤ Y < 0.7) = F (0.7) − F (0.1) ≈
(i) 0.65 (ii) 0.71 (iii) 0.87 (iv) 0.96.
PRGM BETADSTR ENTER ENTER 2 ENTER 5 ENTER 0.7 ENTER subtract PRGM BE-
TADSTR ENTER ENTER 2 ENTER 5 ENTER 0.1 ENTER
2. Beta distribution again: proportion of time car broken. Assume the proportion
of time, Y , a car is broken is approximately a beta with the following density,
(a) What are α and β? Since the general density is given by,
( α−1
y (1−y)β−1
, 0≤y≤1
f (y) = B(α,β)
0, elsewhere,
α − 1 = 0, β − 1 = 4,
(b) What is the chance the car is broken at most 65% of the time?
Since (α, β) = (1, 5), P (Y < 0.65) = F (0.65) ≈ (choose one)
(i) 0.65 (ii) 0.78 (iii) 0.87 (iv) 0.99.
PRGM BETADSTR ENTER ENTER 1 ENTER 5 ENTER 0.65 ENTER
• The moment of random variable Y taken about the origin is defined by,
µ′k = E Y k .
• The moment of random variable Y taken about its mean (or kth central moment
of Y ) is defined by,
µk = E((Y − µ)k ).
12
This is true as long as m(t) exists. The function m(t) exists as long as there is a constant b such
that m(t) < ∞ for |t| ≤ b.
Section 9. Other Expected Values (ATTENDANCE 7) 127
β
R∞
However, since β1∗ 0 e−y/β dy = 1, then
∗
where β ∗ = 1−βt
.
Z ∞
1 ∗ 1 −y/β ∗ 1 1 β
m(t) = × β e dy = × β ∗ = × =
β β 0
∗ β β 1 − βt
1 2 3 4
(i) 1−βt
(ii) 1−βt
(iii) 1−βt
(iv) 1−βt
.
(b) Determine E(Y ); that is, show E(Y ) = β using m(t). First notice
E(Y ) = E Y 1 = µ′1 .
So
d1 m(t)
(1) d {(1 − βt)−1 }
µ′1 = m (0) = = = β(1 − βt)−2 t=0 =
dt1 t=0 dt t=0
E Y 2 = µ′2 .
So
d2 m(t) d2 {(1 − βt)−1 }
(2)
= 2β 2 (1 − βt)−3 t=0
µ′2 = m (0) = =
dt2 t=0 dt2 t=0
(α, β) = (i) (1.5, 2.5) (ii) (1.5, 2.0) (iii) (2.5, 1.0) (iv) (2.5, 1.5),
and µ = αβ = (2.5)(1.5) = (i) 1.00 (ii) 2.75 (iii) 3.00 (iv) 3.75,
and σ 2 = αβ 2 = (circle one)
(i) 5.083 (ii) 5.095 (iii) 5.123 (iv) 5.625,
and so P (0 ≤ Y ≤ 3) ≈ (i) 0.35 (ii) 0.40 (iii) 0.45 (iv) 0.58.
PRGM GAMDSTR ENTER ENTER 2.5 ENTER 1.5 ENTER 3 ENTER.
2
(a) Moment–generating function m(t) = e2t+(1/2)t (2) is normal where
µ = (circle
√ one) (i) 1 (ii) 2 (iii) 3 (iv) 4,
and σ = σ 2 ≈ (circle one) (i) 1.41 (ii) 2.41 (iii) 3.41 (iv) 4.41,
and so P (0 ≤ Y ≤ 3) ≈
(circle one) (i) 0.41 (ii) 0.68 (iii) 0.91 (iv) 0.98.
√
2nd DISTR normalcdf(0,3,2, 2)
2
(b) Moment–generating function m(t) = e2t+(1/2)t is normal where
µ = (circle
√ one) (i) 1 (ii) 2 (iii) 3 (iv) 4,
and σ = σ 2 ≈ (circle one) (i) 1 (ii) 1.41 (iii) 1.51 (iv) 2,
and so P (0 ≤ Y ≤ 3) ≈
(circle one) (i) 0.41 (ii) 0.68 (iii) 0.82 (iv) 0.98.
2nd DISTR normalcdf(0,3,2,1)
etθ2 − etθ1
m(t) = .
t(θ2 − θ1 )
2t t
(a) Moment–generating function m(t) = e −e t
is uniform which is defined on
(θ1 , θ2 ) = (i) (0, 1) (ii) (1, 2) (iii) (−1, 1) (iv) (1, 3),
and µ = θ1 +θ 2
2
= (circle one) (i) 1.0 (ii) 1.5 (iii) 2.0 (iv) 2.5,
θ√
2 −θ1
and σ = 12 ≈ (circle one)
(i) 0.289 (ii) 0.295 (iii) 0.323 (iv) 0.433,
and P (0 ≤ Y ≤ 3) ≈ (circle one) (i) 0.25 (ii) 0.50 (iii) 0.75 (iv) 1.
Since uniform defined on (1,2), probability must be 1.
(i) 5e3t mY (t) (ii) e3t mY (5t) (iii) e3t mW (5t) (iv) e3t mY (5t2 ).
2t t
For uniform moment–generating function, mY (t) = e −e t
,
2. Tchebysheff ’s theorem and gamma: time to fix a car. Suppose the distribution
of the time to fix a car, Y , is gamma, with parameters α = 2, β = 4. Let repair
costs equal C = 25Y + 250.
4.12 Summary
CONTINUOUS f (y) m(t) µ σ2
1 α αβ
Beta B(α,β)
y α−1 (1
− y) β−1
α+β (α+β)2 (α+β+1)
etθ2 −etθ1
Uniform 1/(θ2 − θ1 ) t(θ2 −θ1 )
(θ1 + θ2 )/2 (θ2 − θ1 )2 /12
1
1 −β y 1
Exponential β
e 1−βt
β β2
y α−1 e−y/β 1
Gamma β α Γ(α) (1−βt)α
αβ αβ 2
2 2 µt+σ2 t2 /2
Normal √ 1 e−(y−µ) /2σ e µ σ2
2πσ