Seoul National University Instructor: Junho Song

Dept. of Civil and Environmental Engineering [email protected]

457.212 Statistics for Civil & Environmental Engineers

In-Class Material: Class 10
Probability Distribution Functions and Partial Descriptors (A&T: 3.1)

Example 1: Suppose the likelihood of accidents is uniform along the 100 km highway. Let
X denote the distance between the starting point and an accident location. Determine

0 100

(a) Probability density function (PDF) of X and plot

(b) Cumulative distribution function (CDF) of X and plot

(c) P(20  X  50) by use of PDF and CDF

f X (x)

FX (x)

ex01_sample = runif(10000, min=0, max=100)

# generate random numbers following uniform distribution U(0,100)
hist(ex01_sample, freq=FALSE, breaks=seq(0,100,5), main="PDF")
# plot empirical PDF from samples: use hist with freq=FALSE
plot(ecdf(ex01_sample), verticals=TRUE, pch="", main="CDF")

1
Seoul National University Instructor: Junho Song
Dept. of Civil and Environmental Engineering [email protected]

# plot empirical CDF using samples

ex01_c = subset(ex01_sample, (ex01_sample>=20 & ex01_sample<=50))

# subset of elements satisfying the condition
c_ans = length(ex01_c)/length(ex01_sample) # P(20≤X≤50)

Example 2: Suppose an object can fall anywhere within a 10-km radius circle at random
(i.e., uniform likelihood over points inside the circle). Determine PDF and CDF of the
distance between the location and the center of the circle, 𝑅.

# creating a user-defined function

ex02 = function(num) { # create a function
samp = c()
repeat{ # repeated infinitely without ‘break’
x = runif(1, min=-10, max=10)
y = runif(1, min=-10, max=10)
if(x^2+y^2 < 100){ # random sample is located inside of the circle
samp = rbind(samp, c(x,y))
}
if(dim(samp)[1]==num) break # break when number of sample is ‘num’
}
return(samp) # result of function ‘ex02’ is ‘samp’
}

# scatter plot of samples

ex02_sample = ex02(1000)

2
Seoul National University Instructor: Junho Song
Dept. of Civil and Environmental Engineering [email protected]

plot(ex02_sample, xlab="x", ylab="y", asp=1)

# draw CDF
distance = sqrt(ex02_sample[,1]^2+ex02_sample[,2]^2)
ex02_sample = cbind(ex02_sample, distance)
plot(ecdf(ex02_sample[,3]), verticals=TRUE, pch="", main="CDF")

1. Partial Descriptors of a random variable

(a) “Complete” description by probability functions:

(b) “Partial” descriptors: measures of key characteristics; can derive from ( )

“Numerical descriptions of sample are estimates of partial descriptors representing

populations” e.g., 𝑥̅ and 𝜇

Note:

• Expectation: E[] =  () f X ( x)dx (continuous) or  () p X ( x) (discrete)
− all x

x
𝑛] ∞
∫−∞ 𝑥 𝑛 𝑓𝑋 (𝑥)𝑑𝑥
n
• Moment: 𝐸[𝑋 = or p X ( x)
all x

• Central Moment, E[( X −  X ) ] =  ( x −  X ) n f X ( x)dx or
n
 ( x −  X ) n p X ( x)
− all x

Name Definition Meaning (PDF/CDF)

Location of the ( )
of an area underneath ( )
First moment,
Mean,  X
E[ X ]
Measure of Central Location

The value of a r.v. at which

FX ( x0.5 ) = 0.5 values above and below it
Median, x 0.5 are _______lly probable.
FX−1 (0.5) If symmetric?

The outcome that has the _______est

probability mass or density

Mode, ~
x arg max f X ( x )
x

Becomes more useful as the asymmetry of

the distribution increases

3
Seoul National University Instructor: Junho Song
Dept. of Civil and Environmental Engineering [email protected]

Average of squared deviations

Second-order
central moment
Variance,  2X E[( X −  X ) 2 ]
= E[ X 2 ] − E[ X ] 2
Measure of Dispersion

Radius of ( )

Standard
Deviation,  X  2X

__________ed radius of ( )
Coefficient of X
Variation
(C.O.V.),  X | X |

Third-order central Behavior of two tails

Asymmetry

moment normalized
Coefficient of by  3X , >0
Skewness,  X =0
E[( X −  X ) 3 ] <0
 3X
Fourth-order “Peakedness” - more of the variance is due
central moment to infrequent extreme deviations, as
Flatness

Coefficient of normalized by  4X , opposed to frequent modestly-sized

Kurtosis,  X deviations.
E[( X −  X ) 4 ]
 4X

Example 3: Compute the mean, median, mode, variance, standard deviation and
coefficient of variation for a discrete random variable X whose PMF is given as

x PX (x)
0 0.20
1 0.50
2 0.30

4
Seoul National University Instructor: Junho Song
Dept. of Civil and Environmental Engineering [email protected]

Example 4: Compute the mean, median, mode, variance, standard deviation and coefficient
of variation for a continuous random variable X whose PDF is given by
f X ( x) = (3 / 1000) x 2 , 0  x  10 and 0 elsewhere. What is your guess on the sign of the
coefficient of the skewness? Confirm your guess by computing it.

# mean
ex04_pdf = function(x) {x*(3/1000)*x^2}
mean_data = integrate(ex04_pdf, lower=0, upper=10) # integrate function
mean = mean_data$value # only 'value' of the integrate result

# variance
ex04_var = function(x) {(x-mean)^2*(3/1000)*x^2}
var_data = integrate(ex04_var, lower=0, upper=10)
var = var_data$value
std = sqrt(var)

# skewness
ex04_skew = function(x) {(x-mean)^3*(3/1000)*x^2}

5
Seoul National University Instructor: Junho Song
Dept. of Civil and Environmental Engineering [email protected]

skew_data = integrate(ex04_skew, lower=0, upper=10)

skew = skew_data$value/std^3
# kurtosis
ex04_kurt = function(x) {(x-mean)^4*(3/1000)*x^2}
kurt_data = integrate(ex04_kurt, lower=0, upper=10)
kurt = kurt_data$value/std^4