Random Variables - NCSU Statistics

Document technical information

Format pdf
Size 70.5 kB
First found May 22, 2018

Document content analysis

Category Also themed
Language
English
Type
not defined
Concepts
no text concepts found

Persons

Organizations

Places

Transcript

ST 521: Statistical Theory I
Armin Schwartzman
Class topic 3
Random Variables
Random Variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Conventions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Distribution Functions
Cumulative Distribution Functions .
Some properties of the cdf . . . . . .
Induced Probability Space. . . . . . .
Identically distributed rvs . . . . . . .
Types of Random Variables. . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Discrete random variables
Discrete random variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Properties of the pmf. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Continuous Random Variables
Continuous Random Variabes
Probability density function .
Properties . . . . . . . . . . . . .
Notes . . . . . . . . . . . . . . . .
Notes (cont.) . . . . . . . . . . .
Notes (cont.) . . . . . . . . . . .
Stochastic ordering . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
1
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
2
3
4
5
6
7
8
9
10
11
12
. 13
. 14
. 15
.
.
.
.
.
.
.
16
17
18
19
20
21
22
23
2 / 23
Random Variables
Random Variables
Suppose we start with a probability space (S, A, P ). Instead of referring to outcomes and events observed
from the sample space S, it is often convenient to assign a number to each possible outcome and record
that instead.
Definition: A random variable Y is a real-valued and measurable function defined on a probability
space. That is, Y : S → R.
Every point ω in S maps to a point in R, namely Y (ω).
Conversely, we define the inverse image under Y of a subset B of R as
Y −1 (B) = {ω : Y (ω) ∈ B}
The definition of a random variable requires that the inverse image of every Borel set B ⊂ R is an
element of A. This property allows us to assign probabilities to random variables. More precisely,
P {Y ∈ B} = P {Y −1 (B)}
ST 521 – Class topic 3
Fall 2014 – 3 / 23
Conventions
A Random Variable is a set function which takes values on the real line (for now). Often the argument is
omitted and one writes Y instead of Y (ω).
Random variables are usually denoted by capital letters (e.g. Y ).
Values which random variable can take on are denoted by lower case letters (e.g. y).
Example:
Coin Tosses
S : {H, T }
Y (H) = 1,
Y (T ) = 0
If P (head) = .5, then P (Y = 1) = .5
ST 521 – Class topic 3
Fall 2014 – 4 / 23
2
Examples
Roll of a die:
S = set of sides of a cube
1. Define Y (ω) = # of dots on side ω
P {Y (ω) = ω} = P (ω) = 1/6
2. Define X(ω) = # of dots on side ω modulo 3
ST 521 – Class topic 3
Fall 2014 – 5 / 23
6 / 23
Distribution Functions
Cumulative Distribution Functions
Distribution Functions are used to describe the behavior of a rv.
Definition: The cumulative distribution function (cdf) of a random variable Y is a real valued function
FY (y) defined by
FY (y) = P {Y ≤ y} = P {ω : Y (ω) ≤ y}
Example:
Cdf of a die
Definition: The survival function of Y is defined by
SY (y) = 1 − FY (y) = P (Y > y)
ST 521 – Class topic 3
Fall 2014 – 7 / 23
3
Some properties of the cdf
Let F (y) be a cdf. Then:
1. 0 ≤ F (y) ≤ 1
2. limy→−∞ F (y) = 0
3. limy→∞ F (y) = 1
4. F is nondecreasing: if a < b, then F (a) ≤ F (b)
5. F is right continuous: limy↓b F (y) = F (b)
6. P {a < Y ≤ b} = F (b) − F (a)
These properties can all be proved using the properties of probability measures.
The above properties are also sufficient for F (y) to be a cdf of a rv.
ST 521 – Class topic 3
Fall 2014 – 8 / 23
Induced Probability Space
All probability questions about a random variable can be answered via its cdf.
Every random variable defined on a probability space induces a probability space on R:
(S, A, P ) −→
Y (ω) −→
(R, B, F (·))
Points in S are transformed to points on R (Real line)
Sets (events in A) are mapped into intervals on real line, i.e., into members of the Borel sets, B.
P is replaced by F (·).
Because of this, the abstract notion of a sample space recedes, and attention is usually given primarily to
random variables and their distributions.
We will sometimes refer to the ‘sample space’ of a random variable, which will be taken to be the values
in R that a random variable takes on.
ST 521 – Class topic 3
Fall 2014 – 9 / 23
4
Identically distributed rvs
The cdf does not contain information about the original sample space.
Example: Toss a coin n times. The number of heads and number of tails have the same distribution.
Definition: Two rvs X and Y are identically distributed if for every Borel set A ⊂ R,
P (X ∈ A) = P (Y ∈ A).
Theorem C&B 1.5.10 The following two statements are equivalent:
a. The rvs X and Y are identically distributed
b. FX (x) = FY (x) for every x.
The distinction between two rvs being equal and having the same distribution will become important later
in questions of convergence.
ST 521 – Class topic 3
Fall 2014 – 10 / 23
Types of Random Variables
A random variable Y can be
discrete:
–
–
continuous:
–
–
Y takes on a finite or countably infinite number of values
FY (y) is step-wise constant
the range of Y consists of subsets of the real line.
FY (y) is continuous
mixed: FY (y) is piecewise continuous.
Example: a rv with cdf
⎧
0
⎪
⎪
⎪
⎪
x/2
⎨
F (x) =
2/3
⎪
⎪
11/12
⎪
⎪
⎩
1
ST 521 – Class topic 3
x<0
0≤x<1
1≤x<2
2≤x<3
3≤x
Fall 2014 – 11 / 23
5
12 / 23
Discrete random variables
Discrete random variables
Suppose a random variable Y takes only a finite or countable number of values. Let the sample space of
Y be S = {y1 , y2 , ...}. Then the cdf can be expressed as:
P {Y = yi }
F (y) =
yi ≤y
Definition
The prob. mass function (pmf) or frequency function is a function f (y) defined by
f (y) = P {Y (ω) = y}
If the sample space of Y is S = {y1 , y2 , ...}, then
f (yi ) = P (Y = yi ) = P (yi−1 < Y ≤ yi ) = F (yi ) − F (yi−1 )
Example: Suppose Y is a random variable that takes the values 0, 1 or 2 with probability .5, .3, and .2,
respectively.
ST 521 – Class topic 3
Fall 2014 – 13 / 23
Properties of the pmf
Definition: The domain of a random variable Y is the set of all values of y for which f (y) > 0. This is
also called range or sample space.
Properties of the pmf:
1. f (y) > 0 for at most a countable number of values y. For all other values y, f (y) = 0.
2. Let {y1 , y2 , . . .} denote the domain of Y . Then
∞
f (yi ) = 1
i=1
An obvious consequence is that f (y) ≤ 1 over the domain.
Example: What is the pmf of a deterministic rv (a constant)?
ST 521 – Class topic 3
Fall 2014 – 14 / 23
6
Example
In many applications, a formula can be used to represent the pmf of a random variable. Suppose Y can
take values 1,2 ... with pmf
f (y) =
1
y(y+1)
0
y = 1, 2, ...
otherwise
How would we determine if this is an allowable pmf ?
ST 521 – Class topic 3
Fall 2014 – 15 / 23
16 / 23
Continuous Random Variables
Continuous Random Variabes
A random variable Y is called continuous if its distribution function FY (y) = P (Y ≤ y) is a
continuous function.
A random variable Y is called absolutely continuous if its distribution function F (y) = P (Y ≤ y) is
an absolutely continuous function.
Definition: A function F (y) is absolutely continuous if it can be written
y
f (x)dx.
F (y) =
−∞
Absolute continuity is stronger than continuity but weaker than differentiability.
An example of an absolutely continuous function is one that is:
continuous everywhere
differentiable everywhere, except possibly for a countable number of points.
ST 521 – Class topic 3
Fall 2014 – 17 / 23
7
Probability density function
If F (y) is absolutely continuous, f (y) is called the probability density function (pdf) of Y and
F (y) =
dF (y)
= f (y).
dy
Building on this idea,
P (a < Y ≤ b) = F (b) − F (a) =
a
b
f (x)dx
More generally, for a set B,
P (Y ∈ B) =
B
f (x)dx
Note that of course B has to be an “allowable” subset of the real line R, that is, a Borel set.
ST 521 – Class topic 3
Fall 2014 – 18 / 23
Properties
In general, a function f (x) is a pdf iff
1. f (x) ≥ 0
∞
2. −∞ f (x)dx = 1
Examples:
Suppose F (x) = 1 − e−λx for x > 0 and F (x) = 0 otherwise. Is F (x) a cdf? What is the associated
pdf?
What about f (x) = 1/xr for x > 1 and f (x) = 0 otherwise?
ST 521 – Class topic 3
Fall 2014 – 19 / 23
8
Notes
f (x) is not the probability that Y = x.
In fact, if Y is an absolutely continuous random variable with density function f (x), then
P (Y = x) = 0.
Why?
P (Y = x) =
=
lim
x+h
h→0 x−h
f (u)du
lim F (x + h) − F (x − h)
h→0
= F (x+) − F (x−)
= 0
ST 521 – Class topic 3
Fall 2014 – 20 / 23
Notes (cont.)
More generally, if B is a subset of R with
B
then if Y is an absolutely continuous random variable defined on R, then P (Y ∈ B) = 0 also.
Because P (Y = a) = 0, all the following are equivalent:
P (a ≤ Y ≤ b),
dx = 0,
P (a ≤ Y < b) and
P (a < Y < b)
Also, note that f (x) can exceed one!
ST 521 – Class topic 3
Fall 2014 – 21 / 23
9
Notes (cont.)
f (x) can be interpreted as the relative probability that Y takes the value x. Why? By the mean value
theorem, we can say
P (x < Y ≤ x + Δ) ≈ f (x)Δ
Thus
P (Y ∈ interval of width Δ centered at a) ≈ f (a)Δ
and
P (Y ∈ interval of width Δ centered at b) ≈ f (b)Δ
Hence, if f (b) > f (a), we can say that it is more likely for Y to take the values near b rather than
near a.
ST 521 – Class topic 3
Fall 2014 – 22 / 23
Stochastic ordering
Suppose Y is a rv and define X = Y + 2. Then X > Y always.
Now suppose
X ∼ FX (t) = (1 − e−t ) 1(t > 0)
Y ∼ FY (t) = (1 − e−2t ) 1(t > 0)
Then X is not always greater than Y , but it is likely to be.
Definition: X is stocastically greater than Y if
FX (t) ≤ FY (t) for all t
FX (t) < FY (t) for some t
or equivalently
P (X > t) ≥ P (Y > t) for all t
P (X > t) > P (Y > t) for some t
ST 521 – Class topic 3
Fall 2014 – 23 / 23
10

Similar documents

×

Report this document