Transcript

Beatrice van Eden

Probability DistributionPart 1

Topics

Probability Distribution Probability Distribution Equations Descriptive parameters for Probability Distributions Probability Theorems Binary Variables The beta distribution

Multinomial Variables The Dirichlet distribution

Probability DistributionA function that describes all the possible values and

likelihoods that a random variable can take within a given range.

This range will be between the minimum and maximum possible values, but where the possible value is likely to be plotted on the probability distribution depends on a number of factors, including the distributions mean, standard deviation, skewness and kurtosis.

Probability Distributions Equations

The section on probability equations explains the equations that define probability distributions.

Cumulative distribution function (cdf)Probability mass function (pmf)Probability density function (pdf)

Probability Distributions Equations

Cumulative distribution function (cdf)

The (cumulative) distribution function, or probability distribution function, F(x) is the mathematical equation that describes the probability that a variable X is less that or equal to x, i.e.

F(x) = P(X≤x) for all x

where P(X≤x) means the probability of the event X≤x.

Probability Distributions Equations

Cumulative distribution function for the normal distributions.

Probability density function for several normal distributions. The red line denotes the standard normal distribution.

Probability Distributions Equations

A cumulative distribution function has the following properties: F(x) is always non-decreasing, i.e. F(x) = 0 at x = -∞ or minimum F(x) = 1 at x = ∞ or maximum

( ) 0dF x

dx

Probability Distributions Equations

Probability mass function (pmf)

If a random variable X is discrete, i.e. it may take any of a specific set of n values xi, i = 1 to n, then:

P(X=xi) = p(xi)

p(x) is called the probability mass function

Probability Distributions Equations

The graph of a probability mass function. All the values of this function must be non-negative and sum up to 1.

The probability mass function of a fair die. All the numbers on the die have an equal chance of appearing on top when the die stops rolling.

1 3 7

0.2 0.5 0.3

1 2 3 4 5 6

1/6 1/6 1/6 1/6 1/6 1/6

Probability Distributions Equations

Note that

and F(xk) =

1

( ) 1n

i

i

p x

1

( )k

i

i

p x

Probability Distributions Equations

Probability density function (pdf)

If a random variable X is continuous, i.e. it may take any value within a defined range (or sometimes ranges), the probability of X having any precise value within that range is vanishingly small because a total probability of 1 must be distributed between an infinite number of values. In other words, there is no probability mass associated with any specific allowable value of X.

Probability Distributions Equations

Instead, we define a probability density function f(x) as:

i.e. f(x) is the rate of change (the gradient) of the cumulative distribution function. Since F(x) is always non-decreasing, f(x) is always non-negative.

( ) ( )d

f x F xdx

Probability Distributions Equations

For a continuous distribution we cannot define the probability of observing any exact value. However, we can determine the probability of lying between any two exact values (a, b):where b>a.

( ) ( ) ( )P a x b F b F a

Descriptive parameters for Probability Distributions

The section on probability parameters explains the meaning of standard statistics like mean and variance within the context of probability distributions.

Descriptive parameters for Probability Distributions

Location Mode: is the x-value with the greatest

probability p(x) for a discrete distribution, or the greatest probability density f(x) for a continuous distribution.

Median: is the value that the variable has a 50% probability of exceeding, i.e. F(x50) = 0.5

Descriptive parameters for Probability Distributions

Mean : also known as the expected value, is given by: for discrete variables for continuous variables

The mean is known as the first moment about zero. It can be considered to be the centre of gravity of the distribution.

1

n

i i

i

x p

. ( ).x f x dx

Descriptive parameters for Probability Distributions

SpreadStandard Deviation: measures the amount of

variation or dispersion from the average or mean. The standard deviation is the positive square root of the variance.

The standard deviation has the same dimension as the data, and hence is comparable with deviations of the mean.

Descriptive parameters for Probability Distributions

Variance: measures how far a set of numbers is spread out.

An equivalent measure is the square root of the variance, called the standard deviation.

The variance is one of several descriptors of a probability distribution. In particular, the variance is one of the moments of a distribution.

Descriptive parameters for Probability Distributions

ShapeSkewness:

The skewness statistic is calculated from the following formulae:

Discrete variable:

Continuous variable:

max3

min3

( ) . ( ).x f x dx

S

3

13

( ) .n

i i

i

x pS

Descriptive parameters for Probability Distributions

Kurtosis: The kurtosis statistic is calculated from the following formulae:

Discrete variable:

Continuous variable:max

4

min4

( ) . ( ).x f x dx

K

4

14

( ) .n

i i

i

x pK

Probability TheoremsProbability theorems explains some fundamental probability theorems most often used in modelling risk, and some other mathematical concepts that help us manipulate and explore probabilistic problems.

The strong law of large numbers Central limit theorem Binomial Theorem Bayes theorem

Probability Theorems The strong law of large numbers The strong law of large numbers says that the larger

the sample size (i.e. the greater the number of iterations), the closer their distribution (i.e. the risk analysis output) will be to the theoretical distribution (i.e. the exact distribution of the models output if it could be mathematically derived).

Probability Theorems Central Limit Theorem(CLT)

The distribution of the sum of N i.i.d. randomvariables becomes increasingly Gaussian as Ngrows.Example: N uniform [0,1] random variables.

Probability Theorems Binomial Theorem

a Formula for finding any power of a binomial without multiplying at length.

Properties of binomial coefficient

!

!( )!

n n

x x n x

0

1

10

n

i

n n

n x x

n n n

x x n x

n n

n

a b a b

n b n i

Probability Theorems Bayes theorem

a theorem describing how the conditional probability of each of a set of possible causes for a given observed outcome can be computed from knowledge of the probability of each cause and the conditional probability of the outcome of each cause.

Topics Binary Variables The beta distribution

Multinomial Variables The Dirichlet distribution

Binary VariablesBinary variable Observations (i.e., dependent variables) that occur in one of two possible states, often labelled zero and one. E.g., “improved/not improved” and “completed task/failed to complete task.” Coin flipping: heads=1, tails=0 Bernoulli Distribution

( 1| )p x

1( | ) (1 )

var 1

x xBern x

x

x

Binary VariablesN coin flips

Binomial distribution

( | , )p m heads N

0

2

0

( | , ) ( ) (1 )

( | , )

var[ ] ( [ ]) ( | , ) (1 )

m N m

m

N

m

N

m

Bin m N N

m mBin m N N

m m m Bin m N N

Beta distributionBeta is a continuous distribution defined on the interval of 0 and 1, i.e.,parameterized by two positive parameters a and b.

where T(*) is gamma function. beta is conjugate to the binomial and Bernoulli distributions

0,1

11

2

| , 1

var1

baBetaa b

a ba b

a

a b

ab

a b a b

Beta distribution Illustration of one step of sequential Bayesian

inference. The prior is given by a beta distributionwith parameters a = 2, b = 2, and the likelihood function, given by (2.9) with N = m = 1, corresponds to asingle observation of x = 1, so that the posterior is given by a beta distribution with parameters a = 3, b = 2.

Beta distribution

Example

Beta1.odt

Multinomial DistributionMultinomial distribution is a generalization of the binominal distribution. Different from the binominal distribution, where the RV assumes two outcomes, the RV for multi-nominal distribution can assume k (k>2) possible outcomes.

Let N be the total number of independent trials, mi, i=1,2, ..k, be the number of times outcome i appears. Then, performing N independent trials, the probability that outcome 1 appears m1, outcome 2, appears m2, …,outcome k appears mk times is

Multinomial Distribution

1 211 2

, ,...., | ,...

var 1

cov

KmK

K KKK

K K

K K

j K j K

MultN

m m m Nmm m

m N

m N

m m N

The Dirichlet DistributionThe Dirichlet distribution is a continuous multivariate probability distributions parametrized by a vector of positive reals a. It is the multivariate generalization of the beta distribution.

Conjugate prior for themultinomial distribution.

10

11

1

( | )...

0

K

K

kkK

K

kk

Dir

top related