What Is Probability?

4: Probability
n
N
p
pdf
pmf
RV
x
X
binomial
expected value [parameter]
number of trials [parameter]
normal
probability of success [parameter]
probability density function
probability mass function
random variable
standard deviation [parameter]
value for random variable X (e.g., observed number of successes for a binomial random variable)
random variable X
What is probability?
The probability of an event is its relative frequency in the long run. If an event occurs x times out of n,
then its probability will converge on X n as n becomes infinitely large. For example, if we flip a coin
many times, we expect to see half the flips turn up heads.
Note that when n is small, the observed relative frequency of an event will not be a reliable reflection of
its probability. However, as the number of observations n increases, the observed frequency becomes a
more reliable reflection of the probability. For example, if a coin is flipped 10 times, there is no guarantee
that it will turn up as exactly 5 heads every time. However, if the coin is flipped 10,000 times, chances are
pretty good that the proportions of heads will be pretty close to 0.5.
Random variables (RVs)

A random variable (RV) is a quantity take of various values depending on chance. There are two types
of random variables: (1) discrete random variables and (2) continuous random variables.
Discrete random variables form a countable set of outcomes. We will study binomial random
variables as a way to familiarize ourselves with discrete random variable.
Continuous random variables form an infinite continuum of possible outcomes. We will study
normal (Gaussian) random variables as a way to familiarize ourselves with continuous random
variables.
Page 1 of probability.docx (5/18/2016)
Binomial random variables

Definition
Binomial random variables are discrete RVs of counts that describe the number of successes (X) in n
independent Bernoulli trials, a where each Bernoulli trial has the probability of success p.
Binomial random variables have two parameters:
n the number of independent Bernoulli trials
p the consistent probability of success for each trial
EXAMPLE. As an example of a binomial RV, consider the number of successful treatments (X) in 3
patients (n) where the probability of success in each instance p is 0.25. X can take on the discrete values
of 0, 1, 2, or 3.
Notation. Let b represent the binomial distribution and ~ represent distributed as. Thus, X~b(n, p)
is said random variable X is distributed as a binomial random variable with parameters n and p. The
binomial random variable in our example is X~b(3, .25).
Let Pr(X = x) represent the probability random variable X takes on the value x. For example, Pr(X = 0)
the probability of 0 successes for random variable X. Pr(X x) represent the probability random variable
X takes on a value x or less. This is the cumulative probability of the event.
DEFINITION. The probability mass function (pmf) of a discrete random variable assigns probabilities
for all possible outcomes.
EXAMPLE. The pmf for X~b(3, .25) is shown in Table 1. The exact probabilities of events are shown in
the second column. The cumulative probability is shown in the third column.
TABLE 1. The pmf for X~b(3, .25).
X
Number of successes
0 (event A)
1 (event B)
2 (event C)
3 (event D)
Pr(X = x)
Probability
0.4219
0.4219
0.1406
0.0156
Pr(X x)
Cumulative Probability
0.4219
0.8438
0.9844
1.0000
How we calculated these probabilities is not currently the issue. Instead, let us focus on meaning. The
above pmf states that for X~b(3, .25) we expect to see 0 successes 0.4219 of the time, 1 success 0.4219 of
the time, 2 successes 0.1406 of the time, and 3 successes 0.0156 of the time.
To calculate binomial probabilities, use this app http://www.di-mgt.com.au/binomial-calculator.html. If
you are curious or for some reason you need to calculate binomial probabilities by hand, use the formulas
in Chapter 6 of Basic Biostatistics for Public Health Practice.
A Bernoulli trial is a random event that can take on one of two possible outcomes: success or failure.
Rules for working with probabilities

Let:
A event A (e.g., let A represent 0 success out of 3 in our Example).

B event B (e.g., let B represent 1 success out of 3).
Pr(A) the probability of event A. (e.g., Pr(A) = 0.4129).
the complement of event A not A.
the union of events (e.g., A B A or B).
Rule 1: Probabilities can be no less than 0% and no more than 100%. an event with probability 0 never
occurs and an event with probability 1 always occurs
0 Pr(A) 1
Note that an all the events in Table 1 obey this rule.
Rule 2: All possible outcomes taken together have probability exactly equal to 1.
Pr(all possible outcomes) = 1
Note that in Table 1, Pr(all possible outcomes) = 0.4129 + 0.4129 + .1406 + 0.0156 = 1.
Rule 3: When two events are disjoint (cannot occur together), the probability of their union is the sum of
their individual probabilities.
Pr(A B) = Pr(A) + Pr(B), if A and B are disjoint
In Table 1 let A 0 successes and A 1 success. Pr(A B) = 0.4219 + 0.4219 = 0.8438.
Rule 4: The probability of a complement is equal to 1 minus the probability of the event.
Pr() = 1 Pr(A)
In Table 1, (1, 2, or 3 successes) and Pr() = 1 0.4219 = 0.5781.
The area under the curve (AUC) concept

Probability mass functions (pmfs) can be drawn. For our Example (pmf in Table 1):
X~BINOMIAL(3,.25)
0.45
0.4
0.35
0.3
0.25
0.2
0.15
0.1
0.05
0
0.4219
0.4219
0.1406
X=0
X=1
X=2
0.0156
X=3
Figure 1. X~b(3, .25).
On the horizontal axis, the first bar stretches from 0 to 1. Therefore, this rectangle has base = 1. It has
height = 0.4219. Thus, the area of this bar = h b = 0.4219 1 = 0.4219 = Pr(X = 0).
The second bar also has a base of 1 (from 1 to 2), height of 0.4219, and area = h b = 0.4219 1 =
0.4219. This corresponds to Pr(X = 1).
The combined area of the first two bars = Pr(X = 0) or Pr(X = 1) = 0.4219 + 0.4219 = 0.8438.
In fact the area between any two values is equal to the probability of obtaining an outcome between these
two values. This fact is referred to as the area under the curve (AUC) rule.
We can also use the rule of complements to find AUCs (probabilities).
Let A 0 successes.
Therefore one or more successes.
By the rule of complements, Pr() = 1 0.4219 = 0.5781.
If you add up the Pr(X = 1) + Pr(X = 2) + Pr(X = 3) you will see that this AUC is also equal to
0.5791.
Normal (Gaussian) random variables

Probability density functions (pdf)
We will use normal random variables as a way to introduce continuous random variables.
Probability density functions (pdf) assign probabilities for all possible outcomes for continuous random
variables. A pdf cannot be easily shown in tabular form. I can however be drawn.
While probability functions for discrete random variables (pmfs) are chunky pdfs are smooth curves.
Normal pdf are recognized by their smooth, symmetrical bell shape.
Normal random variables are a family of random variables. Each family member is are characterized by
two parameters, (mu) and (sigma).
the pdfs mean or expected value
the pdfs standard deviation
When changes, the location of the normal pdf changes.
When changes, the spread of the normal pdf changes.
Figure 2. 1 > 2
and are the analogues of and s for data distributions. However, you cannot calculate and . They
describe probability functions, which are distinct from data.
You can get an idea of the size of on a normal pdf by identifying the curves points of inflection. This
is where the curve begins to change slope. Trace the normal curve with your finger. You are skiing
either the slope. The point at which the slope begins to flatten is a point of inflection.
The left inflection point marks the location . This is one -unit below the mean.
The right point of inflection marks the location of . This is one -unit below the mean.
Normal probabilities and the area under the curve (AUC)

Probabilities for continuous random variables are represented as the area under the curve (AUC)see
prior section on AUCs!
The 68-95-99.7 rule helps get a grip on AUCs (probaiblities) from normal random variables. This rule
applies to normal random variables only and says:
68% of the AUC for normal RVs lies in the region

95% of the AUC for normal RVs lies in the region 2
99.7% of the AUC for normal RVs lies in the region 3
Visually, the 95 part of the rule looks like this:
Think in terms of these landmarks:
Although and vary from one normal random variable to the next, you can apply the 68-95-99.7 rule to
any normal random variable. Keep in mind that
(a) AUC = probability
(b) the total AUC = 1
(c) the values lie on the horizontal axis
EXAMPLE. The Wechsler Intelligence Scale is used to measure intelligence. It is calibrated to produce a
Normal distribution with = 100 and = 15 within each age group.
NOTATION. Let X~N(, ) represent a normal random variable with mean and standard deviation .
Using this notation, Wechsler Intelligence scale scores is represented X~N(100, 15).
The 689599.7 rule states that for X~N(100, 15):
68% of the AUC lies in the range = 100 15 = 85 to 115

95% of the AUC lies in the range 2 = 100 (2)(15) = 70 to 130
99.7% of the AUC lies in the range 3 = 100 (3)(15) = 55 to 145
This next figure shows the AUC for X~N(100, 15). Notice the center of the curve is on . Also notice
landmarks at 1, 2, 3 on the horizontal axis.
Finding AUCs with an app for normal random variables

The key concept here is the AUC between any two points corresponds to the probability of that
observation. This applet will calculate AUCs between any two points on any X~N(, ):
http://onlinestatbook.com/2/calculators/normal_dist.html.
Example. Plug in values for X~(100,15). The AUC between 130 and corresponds to the right tail of the
pdf. Note that this AUC (probability) is 0.0228 (roughly 2.5%).

What Is Probability?

Uploaded by

Copyright:

Available Formats

What Is Probability?

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

What Is Probability?

Uploaded by

Copyright:

Available Formats

4: Probability

Random variables (RVs)

Page 1 of probability.docx (5/18/2016)

Binomial random variables

Rules for working with probabilities

A event A (e.g., let A represent 0 success out of 3 in our Example).

the union of events (e.g., A B A or B).

Page 3 of probability.docx (5/18/2016)

The area under the curve (AUC) concept

Figure 1. X~b(3, .25).

Page 4 of probability.docx (5/18/2016)

Normal (Gaussian) random variables

When changes, the spread of the normal pdf changes.

Page 5 of probability.docx (5/18/2016)

Page 6 of probability.docx (5/18/2016)

Normal probabilities and the area under the curve (AUC)

68% of the AUC for normal RVs lies in the region

Visually, the 95 part of the rule looks like this:

Think in terms of these landmarks:

68% of the AUC lies in the range = 100 15 = 85 to 115

Finding AUCs with an app for normal random variables

Page 8 of probability.docx (5/18/2016)

You might also like