Lecture 10
Continuous Random Variables
Part 2
ENIE6x1004 - Probability Theory
[email protected] Today’s discussion
• Gamma Distribution
• Exponential and Chi-Squared Distribution
• Poisson Process
• Other Continuous Distributions
Gamma Distribution
Gamma Distribution (1/6)
The gamma distribution is used to model the time required until the
𝑘 𝑡ℎ event. For example, time until 5 clicks of a website link or time until
12 electronic component failures.
Thus, it may be considered as the continuous counterpart of the
negative binomial distribution. However, time is continuous.
Let 𝑋 be the time required until a certain number of events 𝛼 where
the time between events is known to be 𝛽. Then 𝑋 will have a Gamma
Distribution with parameters 𝛼 and 𝛽.
The key assumption is that each events occurring is independent of the
preceding event, i.e. it follows a Poisson Process.
Gamma Distribution (2/6)
Probability Density Function
The PDF of continuous random variable 𝑋 with a gamma distribution
which parameters are 𝛼 and 𝛽, is
1 𝛼−1 𝑒 −𝑥/𝛽 ,
𝛼
𝑥 𝑥>0
𝑓(𝑥; 𝑎, 𝑏) = ൞𝛽 Γ(𝛼)
0, elsewhere
Where 𝛼 > 0 and 𝛽 > 0, and Γ(𝛼) is the gamma function
∞
Γ 𝛼 = න 𝑥 𝛼−1 𝑒 −𝑥 𝑑𝑥, for 𝛼 > 0
0
Or
Γ 𝛼 = 𝛼−1 !
Gamma Distribution (3/6)
Mean and Variance
𝛼 is called the shape parameter. It is
analogous to the parameter 𝑘 in the
negative binomial which states the
number of “successes”.
𝛽 is called the rate parameter. It is
analogous to the probability parameter 𝑝
in the negative binomial distribution.
The expectation of random variable 𝑋
with a gamma distribution is
𝐸(𝑋) = 𝛼𝛽
And its variance is
𝜎 2 = 𝛼𝛽2
Gamma Distribution (4/6)
Example 1
Suppose that telephone calls arriving at a particular switchboard with
an average of 5 calls coming per minute.
Assuming each calls are independent, what is the probability that up to
1 minute will elapse by the time 2 calls have come into the
switchboard?
Gamma Distribution (5/6)
Example 1
Solution
Since random number of calls are arriving in an interval time (minute) and they are
independent, then this is a Poisson process.
Let 𝑋 be the time in minutes that transpires before 2 calls come, and 5 calls occur
every minute. Then 𝑋 follows a gamma distribution with 𝛼 = 2 and 𝛽 = 1/5.
The probability is given by
1
1 1 −5𝑥
𝑃 𝑋 ≤ 1 = 𝑓 1; 2, =න 2
𝑥𝑒 𝑑𝑥
5 0 (1/5) Γ(2)
𝑥
1 1 𝛼−1 −𝛽 1
= 0 𝛼 𝑥 𝑒 dx = 25 0 𝑥𝑒 −5𝑥 dx
𝛽 𝑟(𝛼)
1 1 1
= 25 − 𝑥𝑒 −5𝑥 - 𝑒 −5𝑥 ሿ ቚ 0
5 25
= 1 - 𝑒 −5 (5+1) = 0.96
Exponential and Chi-Squared
Distribution
Exponential Distribution (1/6)
The exponential distribution is used to predict the time
until the first event occurs. It is a gamma distribution
with 𝛼 = 1. The rate parameter is 𝛽 = 1/𝜆.
If 𝑋 is an exponentially distributed random variable, we
write it as 𝑋 ∼ Expo(𝜆).
Since Γ 1 = 1, the PDF of exponential distribution is
−𝑥𝜆 ,
𝑓 𝑥; 𝛽 = ቊ 𝜆𝑒 𝑥 > 0
0, elsewhere
The corresponding CDF is
𝐹 𝑥 = 1 − 𝑒 −𝜆𝑥 , 𝑥>0
It may be viewed as the continuous counterpart of the
discrete geometric distribution.
Exponential Distribution (2/6)
Let 𝑋~Expo(𝜆), then its expected value is
∞
1
𝐸 𝑋 =න 𝑥𝑒 −𝜆𝑥 𝑑𝑥 =
−∞ 𝜆
And its variance is
2 2 1
Var 𝑋 = 𝐸 𝑋 − 𝐸 𝑋 = 2
𝜆
Exponential Distribution (3/6)
Memoryless Property
07:00 07:10 07:20 07:30 07:40
Suppose you are at a bus stop and have waited for 10 minutes. The bus
arrives but somehow you missed it, and you have to wait for another 𝑠
minutes for the next one.
The memoryless property states that the probability for the bus to
arrive 10 + 𝑠 minutes from the first time you arrive at the stop is equal
to the probability of waiting for the additional time 𝑠.
Exponential Distribution (4/6)
Memoryless Property
Let 𝑋 be a random variable with exponential distribution, then
𝑃 𝑋 ≥ 𝑟 + 𝑠|𝑋 ≥ 𝑟 = 𝑃 𝑋 ≥ 𝑠
For all 𝑟, 𝑠 > 0.
Using the definition of conditional probability, we can verify that
𝑃 𝑋 ≥ 𝑟 + 𝑠 ∩ 𝑃(𝑋 ≥ 𝑟) 𝑃 𝑋 ≥ 𝑟 + 𝑠
𝑃 𝑋 ≥ 𝑟 + 𝑠|𝑋 ≥ 𝑟 = =
𝑃 𝑋≥𝑟 𝑃 𝑋≥𝑟
𝑒 −𝜆(𝑟+𝑠)
= −𝜆(𝑟) = 𝑒 −𝜆𝑠 = 𝑃 𝑋 ≥ 𝑠
𝑒
The Geometric distribution also have this property.
Chi-Squared Distribution
𝑣
A gamma distribution with 𝛼 = and 𝛽 = 2 is called chi-squared
2
distribution. 𝑣 is a positive integer, called the degree of freedom.
Its density function is
1 𝑣
−1 −𝑥/2
𝑣/2
𝑥 𝑒
2 , 𝑥>0
𝑓 𝑥; 𝑣 = ൞2 Γ(𝑣/2)
0, elsewhere
The mean and variance is
𝜇 = 𝑣 and 𝜎 2 = 2𝑣
The chi-squared distribution is an important component of statistical
hypothesis testing and estimation.
Exercise 1 (1/2)
An Internet Service Provider (ISP) has determined that the number of
customer who attempt to connect to the internet per hour has a
Poisson Distribution with 𝜆 = 30 per hour. The time between
connection is exponentially distributed.
The system configuration states that if connections are made under 45
seconds, then the connection fails.
The top management would like to have an availability rate of at least
95%. Does the system configuration satisfy this target?
Exercise 1 (2/2)
Solution
The time between connection is 𝜆 = 30 per hour = 0.5 per minute.
If time between connections made under 45 seconds or 0.75 minutes,
the connection fails.
Then, the probability for two connections made under 0.75 minutes is
0.75
𝑃 𝑋 ≤ 0.75 = න 𝜆𝑒 −𝜆𝑥 𝑑𝑥 = 1 − 𝑒 − 0.5 (0.75)
= 0.3127
0
In other words, 31.27% of the customers will fail to connect. The
system configuration must be improved.
Exercise 2 (1/2)
Fred lives in Blissville, where buses always arrive exactly on time, with the time
between successive buses fixed at 10 minutes. Having lost his watch, he arrives at
the bus stop at a uniformly random time on a certain day (assume that buses run
24 hours a day, every day, and that the time that Fred arrives is independent of the
bus arrival process).
1. What is the distribution of how long Fred has to wait for the next bus? What is
the average time that Fred has to wait?
2. Given that the bus has not yet arrived after 6 minutes, what is the probability
that Fred will have to wait at least 3 more minutes?
3. Fred moves to Blotchville. Now, interarrival of the bus is an Exponential random
variable with mean 10 minutes. Fred arrives at the bus stop at a random time,
not knowing how long ago the previous bus came. What is the distribution of
Fred's waiting time for the next bus? What is the average time that Fred has to
wait?
Exercise 2 (2/2)
a) Let 𝑋 be the time for Fred waiting for a bus. Since the time between
successive buses buses fixed at 10 minutes and Fred arrives at
a uniformly random time, then 𝑋~𝑈 0,10 and its mean is
10 − 0
𝐸 𝑋 = = 5 minutes
2
b) Since 𝑋 is uniformly distributed hence it is not memoryless
1
𝑃 𝑋 ≥ 9∩𝑋 > 6 𝑃 𝑋≥9 10 1
𝑃 𝑋 ≥6+3 𝑋 >6 = = = =
𝑃(𝑋 > 6) 𝑃(𝑋 > 6) 4 4
10
c) Since the distribution is Exponential with parameter 1/10 (and mean 10
minutes), by the memoryless property, how much longer the next bus
will take to arrive is independent of how long ago the previous bus
arrived. The average time that Fred must wait is 10 minutes.
Poisson Process
