Exponential Distribution

Section 8.3 Exponential Distribution

Once again, consider a Poisson Process where you start with an interval of variable length X so that X measures the interval needed in order to obtain a first success with

R = (0, \infty) .

The resulting distribution of X will be called an Exponential distribution.

🔗

To derive the probability function for this distribution, consider finding f(x) by first considering F(x).

🔗

Theorem 8.3.1. Exponential Probability Function.

🔗

For an exponential random variable

X,

f (x) = λ e^{- λ x},

🔗

where

λ

is a given parameter.

🔗

(Note that we will find later that

λ = \frac{1}{μ}

or the reciprocal of the theoretical mean.)

🔗

Theorem 8.3.2. Verification of Exponential Probability Function.

🔗

\int_{0}^{\infty} λ e^{- λ x} d x = 1

🔗

Theorem 8.3.3. Derivation of Statistics for Exponential Distribution and Plotting.

🔗

μ = \frac{1}{λ}

σ^{2} = μ^{2}

γ_{1} = 2

γ_{2} = 9

🔗

Proof.

For the mean, use integration by parts with

u = x

and

d v = λ e^{- λ \cdot x}

to get (eventually)

\begin{aligned} Mean = μ & = \int_{0}^{\infty} x \cdot λ e^{- λ x} d x \\ = [- (x + \frac{1}{λ}) e^{- λ \cdot x}] |_{0}^{\infty} \\ = \frac{1}{λ} . \end{aligned}

and so the use of

λ = \frac{1}{μ}

in f(x) is warranted.

The remaining statistics are derived similarly using repeated integration by parts. The interactive Sage cell below calculates those for you automatically.

🔗

Corollary 8.3.4. Alternate Form for the Exponential Distribution Probability Function.

🔗

Given a Poisson process and a constant

μ = \frac{1}{λ},

suppose

X

measures the variable interval length needed until you get a first success. Then

X

has an exponential distribution with probability function

f (x) = \frac{1}{μ} e^{- \frac{x}{μ}} .

🔗

Theorem 8.3.5. Distribution function for Exponential Distribution.

🔗

F (x) = 1 - e^{- \frac{x}{μ}}

🔗

Example 8.3.6. Router Requests Revisited.

Once again, let’s consider a router which, over time, has been shown to receive on average 1000 such requests in any given 10 minute period during regular working hours. This would mean that, on average, it would take

μ = \frac{10}{1000} = \frac{1}{100} = 0.01

minutes (i.e., less than a second) to receive the first request. If X were to measure the time interval until the first actual request comes in, then the Exponential distribution would be a good model using

f (x) = \frac{1}{0.01} e^{- \frac{x}{0.01}} .

Let’s determine the probability that a first request arrives in the next two seconds. First, note that since X is a continuous variable that f(x) is NOT the probability of exactly X minutes but you must integrate to compute all probabilities. Also, the next 2 seconds is actually the next

\frac{2}{60} = \frac{1}{30}

of a minute. Therefore, F(x) is what you need in general and you find

P (X \leq \frac{1}{30}) = F (\frac{1}{30}) = 1 - e^{- \frac{\frac{1}{30}}{0.01}} \approx 0.96433 .

🔗

Checkpoint 8.3.7. WebWork - Exponential.

Suppose that the time (in hours) required to repair a machine is an exponentially distributed random variable with parameter

λ = 0.8 .

What is

(a) the probability that a repair time exceeds

10

hours?

(b) the conditional probability that a repair takes at least

9

hours, given that it takes more than

6

hours?

Answer 1.

0.000335462627902512

Answer 2.

0.0907179532894125


    
        
xxxxxxxxxx
 
1
r=1                # the number of successes desired
2
            # the mean till first must be given
3
mu = 3
4
sdev = sqrt(mu)  # the formula for the standard deviation
5
6
M = mu*3   # the space is infinite but we just go out 3 standard deviations
7
X = 0:M    # quantiles for the space R of the random variable 
8
9
Pexp <- function(x){dexp(x, mu )}  # create the probability function over X
10
11
curve(Pexp, from=0, to=M, xlab="X", col="blue", lwd=3,
12
 main="Gamma Sampling vs Gamma Curve vs Approximating 'Bell Curve'") 
13
Pnormal <- function(X){dnorm(X, mean=mu, sd=sdev)}   # to overlap a bell curve
14
curve(Pnormal, col="red", lwd=2, add=TRUE) 
15
16
Psample = rexp(10^6, mu)  # to create a histogram, sample a lot
17
# Xtop=max(Psample)          # for scaling the x-axis. Shift by 1/2 below.
18
hist(Psample, prob=TRUE, add=TRUE)

    
    
    
    
        
            
                Language:
                
            
        
    
    




    
    
        
        Messages

🔗

Theorem 8.3.8. The Exponential Distribution yields a continuous memoryless model..

🔗

If X has an exponential distribution and a and b are nonnegative integers, then

P (X > a + b | X > b) = P (X > a)

🔗

Proof.

Using the definition of conditional probability,

\begin{aligned} P (X > a + b | X > b) & = P (X > a + b \cap X > b) P (X > b) \\ = P (X > a + b) / P (X > b) \\ = e^{- (a + b) / μ} / e^{- b / μ} \\ = e^{- a / μ} \\ = P (X > a) \end{aligned}

🔗

Checkpoint 8.3.9.

It can be presumed that the life span in years of a certain brand of lightbulbs follows the Poisson process assumptions. Suppose that the mean life span till failure is known to be 0.7 years. A tester makes random observations of the life times of this particular brand of lightbulbs and records them one by one as either a success if the life time exceeds 1 year, or as a failure otherwise.

Determine the probability that a randomly tested bulb lasts more than 1 year:

Determine probability that the first success occurs in the fifth observation:

Determine the probability that the second success occurs in the 8th observation given that the first success occurred in the 3rd observation:

Determine the probability that the first success occurs in an odd-numbered observation:

Hint.

Use the exponential distribution to get the first answer. Use the geometric to get the remainder.

Answer 1.

0.2397

Answer 2.

0.0801

Answer 3.

0.0801

Answer 4.

0.5681

Solution.

Using the exponential distribution,

p = P (X > 1) = 1 - F (1) = 1 - (1 - e^{- \frac{1}{0.7}}) \approx 0.2397 .

Now, use the geometric distribution with

p = 0.2397 ...

P (X = 5) = f (5) = (1 - 0.2397)^{4} \cdot 0.2397 \approx 0.0801 .

P (2nd success on 8 | 1st success on 3) = P (X = 5) \approx 0.0801

\begin{aligned} P (1st success on an odd trial) & = f (1) + f (3) + f (5) + . . . \\ = p + (1 - p)^{2} \cdot p + (1 - p)^{4} \cdot p + (1 - p)^{6} \cdot p + . . . \\ = p \sum_{x = 0}^{\infty} {((1 - p)^{2})}^{k} \\ = p \frac{1}{1 - (1 - p)^{2}} \approx 0.5681 \end{aligned}

Essentials of Mathematical Probability and Statistics

Search Results:

Section 8.3 Exponential Distribution

Theorem 8.3.1. Exponential Probability Function.

Proof.

Theorem 8.3.2. Verification of Exponential Probability Function.

Proof.

Theorem 8.3.3. Derivation of Statistics for Exponential Distribution and Plotting.

Proof.

Corollary 8.3.4. Alternate Form for the Exponential Distribution Probability Function.

Proof.

Theorem 8.3.5. Distribution function for Exponential Distribution.

Proof.

Example 8.3.6. Router Requests Revisited.

Checkpoint 8.3.7. WebWork - Exponential.

Theorem 8.3.8. The Exponential Distribution yields a continuous memoryless model..

Proof.

Checkpoint 8.3.9.