Blog Logo
Research Scientist @ Foundation AI
·
read
Image Source: https://myplanning.ca/wp-content/uploads/2016/09/dice-game-black-close-up.jpg
· · ·

Probability Series

Basic Probability Concepts

Conditional Probability and Bayes’ Rule

Discrete Random Variables

Continuous Random Variables

· · ·

Discrete Random Variables

Even though it is named variable, discrete random variable is actually a function that maps the sample space to a set of discrete real values.

  • Where
    • X is the random variable
    • S is the sample space
    • R is the set of real numbers

Probability Mass Function

Probability mass function is the probability defined over a given random variable, i.e., it gives the probability that a discrete random variable is exactly equal to some value in the sample space, S.

  • For a random variable X, \(p(X=c)\) is denoted as \(p(c)\) and the mapping of each value in sample space to their respective probabilities is known as pmf.
  • For all values c that are not is sample space \(p(c) = 0\) because it is pointing to an empty set.

Commutative Distribution Function

Probability defined over an inequality such as \(X \leq c\) gives probabilities of all the events that satisfy the condition from \(-\infty\) to c i.e. the probability value estimate for X less than or equal to c. Mathematically,

Explaination

  • Set has 5 boys and 5 girls.
  • 3 kids selected at random but gender not known.
  • X is the random variable that denotes number of girls selected.
  • Event c such that \(X(c) = 3\) is given by set \(c = \{(GGG)\}\) i.e. all three girls selected are girls.
  • The pmf for random variable X will be as follows:
a p(a) CD(a)
0 1/12 1/12
1 5/12 6/12
2 5/12 11/12
3 1/12 1
  • Calculations:
    • Number of ways of selecting k kids from total of N kids is given by \(C_k^N = \frac{N!}{k! (N-k)!}\) implies \(C_3^{10} = \frac{10!}{3! (10-3)!} = 120\) for N = 10 and k = 3.
    • value 1 : a = 0 means that no girl is selected which implies that k boys are selected out of the total B boys. The number of ways this can be accomplished is given by \(C_k^B = \frac{B!}{k! (B-k)!}\) implies \(C_3^5 = \frac{5!}{3! (5-3)!} = 10\) for B = 5 and k = 3.
    • For commutative distribution \(CD(2) = p(0) + p(1) + p(2) = 11/12\).
    • Similarly the value of \(p(a)\) and \(CD(a)\) at other values of \(a\) can be calculated.

Random Variable Metrics

Expected Value

The average or mean value calculated over all the possible outcomes of the random variable.

\[E(X) = \sum_n v_i * p(v_i) \tag{3}\]

  • X is the random variable.
  • \(v_i\) is the value the random variable takes with probability \(p(v_i)\).
  • sample space size is n.
  • often represented by \(\mu\).
  • it is a measure of central tendency of the random variable.
  • Some other properties of expected value of a random variable:

Variance and Standard Deviation

Variance gives the dispersion of probability mass around the mean value (i.e. E(X), expected value) of the random variable.

\[Var(X) = E((X-\mu)^2) \tag{6}\] \[\sigma = \sqrt (Var(X)) \tag{7}\]

  • \(\sigma\) is the standard deviation.
  • Some other properties of variance of a random variable:

\[Var(X) = E(X^2) - (E(X))^2 \label{8} \tag{8}\] \[Var(aX + b) = a^2 Var(X) \label{9} \tag{9}\] \[Var(X+Y) = Var(X) + Var(Y) \text{ iff X and Y are independent } \tag{10}\]

Derivation of equation \eqref{8}

Derivation of equation \eqref{9} using equations \eqref{5}, \eqref{8}

Some Specific Distributions

Uniform Distribution

  • All outcomes are equally possible.
  • Eg: Probability of getting a heads or tails for a fair coin.
  • Uniform(N) implies N outcomes and each has probability 1/N.

Bernoulli Distribution

  • Used to model the random experiment where each trail has exactly 2 possible outcomes.
  • One possible outcome is termed as success and other as failure.
  • Parameter \(p\) is the probability of success in an experiment.
  • Random variable \(X \in \{0, 1\}\).
  • X = 1 has probability \(p\) and X = 0 has probability \(1-p\) where 0 is denoting failure and 1 denoting success.

Binomial Distribution

  • Used to model \(n\) independent trails of Bernoulli Distribution.
  • If X follows Binomial(n, p) then, X = k implies, the event of k successes in n independent Bernoulli trials.

\[P(X=k) = C_k^n * (p^k) * ((1-p)^{n-k})\]

  • \(C_k^n\) is the number of ways of picking k success events in n total trials.
  • \((p^k) * ((1-p)^{n-k})\) gives the combined probability of the n Bernoulli trails.

Geometric Distribution

  • Also defined over Bernoulli distribution.
  • Models the event of k failures before first success.
  • \(geometric(p)\) is given by

\[P(X=k) = (1-p)^k * p\]

where \(X = k\) is the event where first success occured after k failures.

  • If X and Y follows geometric distribution with same probability p, then \(X + Y\) is also a geometric distribution.

Expected Value and Variance for Distributions

Distribution E(X) Var(X)
Uniform \(\frac{n+1}{2}\) \(\frac{n^2-1}{12}\)
Bernoulli \(p\) \(p(1-p)\)
Binomial \(np\) \(np(1-p)\)
Geometric \(\frac{1-p}{p}\) \(\frac{1-p}{p^2}\)

REFERENCES:

Discrete Random Variables
Expectation and Variance
Variance

· · ·