Professional Documents
Culture Documents
Distribution
The Poisson Distribution is a discrete probability distribution that applies to occurrences of some event
in a specific interval. The interval can be time, distance, area, volume, etc.
Poisson Probability function
The Poisson distribution is the binomial distribution, except for the fact that it gives the probability of
occurrence over a specified interval in time or space. Thus, the
Poisson distribution can be derived from the binomial
distribution:
1. The expectation value (mean) over a fixed interval is
given,
e.g. on the average 96 large earthquakes occur every 100
years. (μ=96)
2. Divide the interval (100 years) into n small subintervals of
equal length. If the interval is small enough, the event
either occurs once or does not occur at all. Then, the
distribution of the number of occurrences should be
described by the binomial distribution.
3. Let us denote the probability of occurrence at each subinterval p , then
n
n x
x
n!
P( x ) n Cx p x (1 p)n x 1
(n x )! x ! n n
Taking n→∞, hence very small p, to represent a continuous interval, we obtain the Poisson
distribution
xe
P( x )
x!
Poisson mean (expected value) 𝜇 𝐸 𝑋 𝜇
Poisson Standard deviation
Page | 38
Example: Crushed ore entering a processing plant on a continuous conveyor is inspected to ensure it
does not contain any pieces that are too large, as they may damage the processing equipment. The
number of pieces found to be too large occurs randomly, and during a 5 minute period an average of 3.4
large pieces are found.
a) What is the probability that during a 5‐minute period an inspector will find more than 3 pieces
that are too large?
Page | 39
b) Suppose the process needs to be shut down if more than 15 oversize pieces are detected in a 15
minute period. What is the probability that the process will be shut down in any given 15 minute
period?
c) During a routine inspection, 6 pieces of oversize ore are found in a 5 minute period. Is this an
unusual result?
Page | 40
Continuous Probability Distributions
Consider histograms of a large (n∼∞) popula on with different class sizes (e.g. students’ height
distribution), where n is the number of classes:
If we take the limit that the number of classes n goes to infinity, the distribution becomes continuous,
and the probability of each class approaches 0, i.e., P(x)→0.
For continuous distributions
The height of the curve is NOT probability but instead something called
probability density.
The shape of the distribution is given by the probability distribution function:
the PDF
All probabilities are found by calculating areas under the PDF
Single‐point probabilities are zero
Example : Many computer languages have a random number generator that follows a Uniform
distribution. Assume we have such a generator for finding a random number between 0 and 10. Find the
probability of getting a random x‐result between x=1.5 and x=2.3 is?
PDF: f(x)=0.1
Simulation with R:
Page | 41
Probability given a PDF
For a given probability density function f ( x ) , the probability that a x b is
b
P ( a x b) area under the pdf between a and b = pdf ( x)dx
a
The last box is an integral but don’t panic if you have never done calculus we will not be using this. We
can get integration results from Wolfram Alpha.
Area under curve equals probability
Page | 42
The normal distribution (or the Gaussian distribution) is the most commonly used distribution because
• The sum of a large number of random variables is approximately normally distributed.
• It is a good approximation to some discrete distributions, and easier to handle mathematically.
The probability density function of the normal distribution is
2
1 x
1
2
pdf ( x) e
2 2
where μ is the mean, and σ is the standard deviation.
This function is very hard to integrate, therefore we need tables of
values or help from technology (in this case, R) to do so.
The functions dnorm, pnorm, qnorm, and rnorm will provide
solutions to most of our needs related to normal distribution. We will
see examples of them soon.
Properties of the normal distribution
• Mean = Median = Mode, that is, symmetrical about the mean
• Inflection points at x=μ±σ
• The empirical rule is exact.
R code for plotting normal distribution:
R code for plotting cumulative normal distribution (Sigmoid function)
Page | 43
Example : The measured values of atmospheric CO2 concentration is
normally distributed with the mean of 349ppm and the standard
deviation of 24ppm. Calculate the probability that a given measurement
is within 349±24 ppm.
What is probability of getting a CO2 level below 2 ppm?
Example: Suppose that IQ scores are known to be normally distributed with a mean of 100 and a standard
deviation of 15. What is the probability of randomly selecting an IQ score that is greater than 135?
Reversed Example: For a medical study, a researcher wishes to select people in the middle 60% of the
population based on blood pressure. If the mean systolic blood pressure is 120 and the standard deviation
is 8, find the upper and lower readings that would qualify people to participate in the study.
Page | 44
Example: A study of life expectancy after coronary artery bypass graft (CABG) surgery found that the
mean life expectancy was 13.9 days if the subject suffered from impaired left ventricular ejection
fraction (LVEF). The standard deviation for this population is 2.1 days. Assuming the mean and standard
deviation are correct for the total population of patients with CABG and impaired LVEF, and assuming
this population is normally distributed answer the following questions
1. How many people in the population would have a life expectancy greater than 16 years?
2. What is the life expectancy of the top 10% of this population?
The Standard Normal Distribution
x
Using the z‐score ( z ) instead of the raw random variable x, the PDF in terms of z is
1 12 z 2
pdf ( z ) e
2
which has the mean of 0 and the standard deviation of 1.
The standardized normal variable (z‐score) can be helpful when
solving problems with unknown 𝜇 or 𝜎
Page | 45
Examples of using R to get values for probability distributions:
Binomial distribution:
The binomial distribution function is given by: 𝑃 𝑋 𝑥 𝑓 𝑥 𝐶 𝑛, 𝑥 . 𝑝 . 1 𝑝
Poisson distribution:
The PDF is given by: 𝑓 𝑥
!
If 𝑋 𝑃𝑜𝑖𝑠𝑠 𝜆 12 :
Normal distribution:
^
PDF of a normal distribution is given by: 𝑓 𝑥 𝑒
√
If 𝑥 𝑁 5,2
If the mean and sd are not passed, they are assumed to be zero and one respectively
Page | 46
Sampling Distributions
Sampling a population is done to infer properties of the underlying population. A random sample of size
n is a set of n objects drawn from a population in such a way that any set of size n has an equal chance of
being drawn. In this course, we will study the sampling distribution of the mean and standard deviation.
The diagram below compares the sampling distribution of the mean for three different probability
distributions. The top row shows the underlying population’s probability distribution. The rest of the rows
show the sampling distribution of the sample mean and how it changes as the sample size increases.
Page | 47
The Central Limit Theorem (CLT)
This theorem states that if random samples of size n are drawn from any population with mean μ and
standard deviation σ, then, when n is large:
• The distribution of the sample means is approximately a normal distribution. This
approximation becomes better as n gets larger. (Note that if the population is normally
distributed to begin with, then so is the sampling distribution, regardless of sample size.)
• The mean of the sampling distribution of the sample means x is the same as the mean of the
underlying population μ.: x
• The standard deviation of the sampling distribution of the sample means (standard error) is
x , where σ is the standard deviation of the underlying population. (Note that x is
n
inversely proportional to n . This explains why large samples are more accurate than small
samples.)
Sampling distribution is approximately normal when
1. the sample size is large enough
When calculating probability we need the standard deviation σ of the population. If it is
known, say from previous sampling, then we should use it. If it is not known, then for large
samples we can simply replace it with s. The CLT still applies.
2. or the underlying population is normally distributed.
The CLT does not apply when the sample is small and σ is not known. In this case, another
distribution, the Student t distribution, may be used.
Example: Men have weights that are normally distributed with a mean of 172 lb and a standard
deviation of 29lb.
a) Find the probability that if an individual man is randomly selected; his weight will be less than
167 lb.
Page | 48
b) Find the probability that 36 randomly selected men will have a mean weight that is less than 167 lb.
c) If 4 men are randomly selected, find the probability that they have a mean weight between 160 lb and
180 lb.
Find the probability that if five men are selected at random, at least two of them will weigh less than
167 lb.
Page | 49