Business Inferential Statistics

By

G. S. Msukwa

2023
Lesson One

Probability Distributions

A probability distribution is a mathematical function that gives the probabilities of occurrences

of different possible outcomes for an experiment. It is a frequency distribution of probabilities
of occurrences of any variable x, in place of numerical frequencies.
For instance, in an experiment with 4 possible outcomes, and we conduct 10 trials which one
outcome occurs five times, the second outcome occurs three times, the third outcome occurs
two time, and the fourth outcome occurs once. The results can be presented in a frequency
distribution, or the outcomes can be converted into probabilities and a probability distribution
be presented instead.

The types of distributions are identified by their skewness. That is, some distributions are
symmetrical while others are skewed either ways (asymmetrical). These distributions fall under
two basic categories depending on their underlying variables, namely; discrete random
probability distributions and continuous random probability distributions.

1.1 Discrete Random Probability Distributions

A random variable is a numerical description of an outcome of an experiment.

Discrete variables are those variables which only assume finite sequenced numbers (ie, 5
people). Continuous variables are those that may take any value, including fractions of a unit
(ie, measuring weight; 2.5 kg, length etc.)

Probability distributions of discrete random variables can be presented in a frequency table and
frequency charts to show the relative strength or possibility of occurrence of particular
outcomes or values. In this case we consider two most relevant discrete random probability
distributions in economic and business setup.

Binomial Distributions

Firstly, let us consider a class of experiments which meet the following conditions. A binomial
distribution are outcomes of an experiment that satisfies all the four conditions below;
1. The experiment consists of (n) identical trials,
2. Two outcomes are possible on each trial (composed of a success and a failure).
3. The probability of the two outcomes do not change from one trial to the next.
4. The trials are independent (ie, the outcome of one trial does not affect the outcome of any
other trials).
Polynomials (multinomials) are distributions with specified number of trials (n), but where
more than two outcomes are possible from each experiment trial. The probability of the
outcomes do not change. The trials are also independent.
A Bernouli process is an experiment which satisfies condition: 2, 3 and 4 above.

A mathematical binomial probability of a discrete random variable, such as a success of (x),

taking any values; 0, 1, 2, ………n, is given by;

f(x) = n! Px (1 – p)n – x
x!(n – x)!

Where; n = number of trials

P = probability of success in one trial
X = the number of successes in (n) trials
F(x) is the probability of success in (n) trials

That’s, the probability of obtaining value 0 or 1 or 2 etc, from any given trial is f(0), or f(1) or
f(2), respectively

Expected values and variance of the Binomial distribution are given as follows;

E(x) = µ = ∑xf(x) = np, (also refer to Binomial Probability table)

Var(x) = σ2 = np(1 – p)

Note that the Binomial distributions is positively skewed.

Binomial tables can also be used to check up the f(x) given relevant information.

Poisson Distributions

In the Poisson probability distribution, if we consider a number of occurrences of an event over

a specified interval of time or space (ie, a number of road accidents per hour, a number of
phone enquiries received per day, etc), and the following conditions are satisfied;
1. The probability of occurrence of the event is the same for any two intervals of equal length,
2. The occurrence/non-occurrence of the event in any interval is independent of the
occurrence/non-occurrence in any other interval,

then the occurrence follows a Poisson distribution where the probability of a Poisson random
variable is given by:

f(x) = λxe-λ
x!
here, x = number of occurrences in the interval
λ = mean or averages number of occurrences in an interval
 e= the exponent (2.71828)
F(x) the probability of (x) occurrences in the interval

Expected values and variance of the Poisson distribution are given as follows;

E(x) = µ = λ
Var(x) = σ2 = λ

Note that the Poisson Distributions is also positively skewed.

Illustrative examples;

1.2 Continuous Random Probability Distributions;

These distributions are for variables which can take any values. For instance, length can be
measured continuously including fractions of units chosen (ie, 1.85m , 3.78km, etc). For this
reason the distributions are presented as continuous curves (the probability density function -
pdf ). The probability of occurrence of any value below a given value, above the value, or
between any two values is represented by area under the curve for that range.

The relevant distributions under considerations include the following;

o The normal probability distributions
o The exponential probability distributions

The Normal Distribution

This is used to describe a continuous random variable whose probability distribution or

probability density function (pdf) is symmetric, or is not skewed to any side (ie, has a bell
shape), as shown in the graph below:

In this distribution the pdf (area under the curve) follows;

1 −¿ ¿
f (x) = e
σ √2 π
Where; x = the value of one of the items in the distribution
σ2 = the variance of variable (x)
σ = the standard deviation of variable (x)
∏ = the pi (3.14159)
e = the exponent (2.71828)
µ = the mean or expected value 0f the random variable (x)
f(x) is the probability of (x) value in the random variables.
Note;
The size of the standard deviation (σ) around the mean (µ) affects the shape of the of the
normal distribution curve (ie, larger σ means a broader shape)

A standard normal distribution has the following characteristics;

1. σ = 1
2. µ = 0
3. 68.26 of the (x) are within 1 σ
95.44 of the (x) are within 2 σ
99.72 of the (x) are within 3 σ

4. Z = X-µ
σ
2
−z
f(z) = 1 e 2
√2 π
Note; use graphs and standard normal tables for determination of the probabilities (ie, the
area under the curve)

Some other normal probability distributions which do not fall under the standard normal
category may fall under the (student) t-distributions. Although the t-distribution is also
symmetrical the various shapes/sizes depend on the degrees of freedom (df). As the df
increase, the difference between the t-distribution and standard normal becomes smaller, since
the standard normal distribution is assumed to have infinite degrees of freedom (df = ∞ )

Exponential Distribution

This is a continuous distribution of time between events which occur continuously and
independently at a constants average rate. The pdf decreases at a constant rate with x values
(ie, service time). The probability of one variable x is the area under the curve corresponding to
an interval that variable assumes. See graph below;

In this distribution the pdf (area under the curve) follows;

−x
1 μ
f(x) = e for; x ≥ 0, µ ˃ 0
μ
−x
p ( x ≤ x 0) = 1−e μ
0

Expected values and variance of the Binomial distribution are given as follows;
E(x) = µ = 1/µ
Var(x) = σ2 = 1/µ2

Illustrative examples
1.3 Other distributions

The Chi-square (X2) Distribution

This is a continuous distribution of the sum of squares of random variables from samples.

A X2 distribution with k degrees of freedom (df) is the distribution of the sum of squares of k
independent standard normal random variables; ref, graph below:

In this distribution the pdf or the probability of any random variable x is given as;

k −x
2−1 2
x e
f (x) = k
k , for x ≥ 0
2 ⎾( )
2
2

Or f(x) = 0, otherwise

Where; f (x) is the probability density function (pdf)

k is the degrees of freedom
⎾
k
2 ()
is the gamma function
The gamma function is an extension of the factorial function to complex numbers.

Rather than other distributions above, chi-square (X2) is mainly used as a test statistic for
inference tests of population variances from sample variances of a normally distributed
population of variables. It is also used as a test statistic for independence, test of goodness of fit
in multinomial, Poisson and normal distributions.

If a sample variance (s2) is a point estimator of the population variance (σ 2), in using the (s2) as
a basis for making inference about the population variance (σ 2), the sampling distribution of the
quantity (n – 1) s2/ σ2 is important.

When a simple random sample n is selected from a normally distributed population, the
sampling distribution of (n – 1)s2 , has a x2 distribution with (n – 1) degrees of freedom.
σ2
In other words, if z1, z2 ……zk are independent standard normal variables, then the sum of their
squares is distributed according to x2 with kdf.

Ie, Σz2 ʅ x2k (sum of z2 is distributed as X2 with k degrees of freedom –df.)

As the degrees of freedom increase the X2 approaches the normal distribution, such that, with k
˃ 50, the distribution is sufficiently close the normal distribution (becomes asymptotic). We
normally use X2 tables to check up the probability values for the tests.

The F distribution

The F distribution (after Ronald Fisher), is also a continuous probability distribution which arises
usually as a null distribution as a test statistic in analysis of variance (ANOVA) and other F tests.
It is a ratio between two estimates of an index. In ANOVA, it is a ratio of two population sample
variance. It is also used in regression analysis.

The pdf for a random variable x ʅ F(n1, n2), is given by;

f(x) = Г(n1 + n2) Г(n1/2) Г(n2/2)

2
Where, n1 = df for numerator,
n2 = df for denominator

Refer also to the graphical presentation below and the illustration of the area under the curve
for α 0.05.

As the degrees of freedom increase the F distribution approaches the normal distribution. At df1
= 100, df2 = 200, the F distribution is sufficiently close to the normal distribution (becomes
asymptotic). We normally use F distribution tables to check up the probability values for
relevant tests.