STANDARD PROBABILITY DISTRIBUTIONS

Comprehensive Structural & Applied Reference Guide for UGC NET Statistics
Subject: Statistics (Code 107) | Unit I: Probability and Distributions

This master framework compiles the theoretical structures, core operational boundaries, and direct real-world application
diagnostics for standard univariate distributions. It integrates rigorous algebraic formulations with precise diagnostic rules
needed to identify and fit distribution models under exam and applied conditions.

PART 1: STANDARD DISCRETE UNIVARIATE DISTRIBUTIONS

Discrete distributions govern categorical or count data where the sample space contains a finite or countably infinite sequence
of distinct outcomes.

1. Discrete Uniform Distribution

Models a system where a finite set of clean, isolated outcomes all share an identical, flat probability profile across the entire
support.

PMF: P(X = x) = 1 / n, x ∈ {x1, x2, ..., xn}

Parameters: n ∈ &mathbb{N} (The total number of possible distinct outcomes)

2. Bernoulli Distribution

The foundational component of discrete probability networks, capturing a single isolated trial with exactly two mutually
exclusive outcomes: success (1) or failure (0).

PMF: P(X = x) = px (1 - p)1 - x, x ∈ {0, 1}

Parameters: p (Probability of success, where 0 ≤ p ≤ 1)
Moments: E[X] = p, Var(X) = p(1 - p)
MGF: MX(t) = (1 - p) + pet = q + pet (where q = 1 - p)

3. Binomial Distribution — Bin(n, p)

Represents the total count of accumulated successes achieved across a fixed number of completely independent and identical
Bernoulli trials.

PMF: P(X = x) = nCx px qn - x, x = 0, 1, 2, ..., n

Parameters: n ∈ &mathbb{N} (Total trials), p (Success probability, 0 ≤ p ≤ 1)
Moments: E[X] = np, Var(X) = npq
MGF: MX(t) = (q + pet)n

UGC NET English - Statistics (Code 107) Page 1 of 7

 HIGH-YIELD EXAM PROPERTIES

• Additive / Reproductive Property: If X ∼ Bin(n1, p) and Y ∼ Bin(n2, p) are independent variables, then (X + Y)
∼ Bin(n1 + n2, p). Success probabilities (p) must match exactly.

• Shape Profile: Symmetrical if p = 0.5. Displays a positive skew when p < 0.5 and a negative skew when p > 0.5.
• Mean vs. Variance Constraint: The theoretical framework mandates that Mean > Variance (np > npq since 0 < q
< 1).

APPLIED FIELD DIAGNOSTICS & IMPLEMENTATION

• Prerequisites: Fixed number of trials (n), binary categorization, constant success probability (p), and strict trial
independence.
• Real-World Context: Assessing batch quality checks where precisely 100 components are sampled daily,
monitoring the count of items flagged defective.
• Data Diagnostics: Compute sample mean (x̄) and sample variance (s2). If x̄ > s2, the dataset satisfies a basic
binomial requirement. Parameter estimation follows via p̂ = x̄ / n.

4. Poisson Distribution — Pois(λ)

Governs the count of rare, independent random occurrences manifesting uniformly within a fixed interval of continuous time,
area, or space.

PMF: P(X = x) = (e-λ λx) / x!, x = 0, 1, 2, ...

Parameters: λ > 0 (The average rate of occurrence over an interval)
Moments: E[X] = λ, Var(X) = λ
MGF: MX(t) = e λ(et - 1)

HIGH-YIELD EXAM PROPERTIES

• Equidistribution Property: Signature structural feature where Mean = Variance = λ. Frequently targeted in model
identification questions.
• Reproductive Operator: If variables are independent, then (X + Y) ∼ Pois(λ1 + λ2).

• Limiting Form: The Binomial distribution converges to a Poisson distribution as n → ∞ and p → 0 while holding
the product np = λ constant.

APPLIED FIELD DIAGNOSTICS & IMPLEMENTATION

• Prerequisites: Interval independence, arrival rate homogeneity across time, and a zero probability of simultaneous
arrivals at any exact instant.
• Real-World Context: Counting severe industrial accidents per month, or monitoring incoming network traffic
spikes per minute on a web server.
• Data Diagnostics: Verify if sample mean and variance are approximately equal (x̄ &inapprox; s2). The Maximum
Likelihood Estimator (MLE) directly assigns λ̂ = x̄.

UGC NET English - Statistics (Code 107) Page 2 of 7

5. Geometric Distribution

Models the total number of failure events encountered immediately preceding the very first success achieved in a series of
independent Bernoulli trials.

PMF: P(X = x) = qx p, x = 0, 1, 2, ... (where q = 1 - p)

Moments: E[X] = q / p, Var(X) = q / p2
MGF: MX(t) = p / (1 - qet)

HIGH-YIELD EXAM PROPERTIES

• Memoryless Property: The unique discrete distribution displaying the memoryless property: P(X ≥ s + t | X ≥ s) =
P(X ≥ t). Witnessing consecutive failures provides no predictive information about when the next success will appear.

APPLIED FIELD DIAGNOSTICS & IMPLEMENTATION

• Real-World Context: Modelling the count of valid financial transactions audited prior to discovering the first
fraudulent entry.
• Data Diagnostics: Raw sample histograms show an immediate geometric peak at zero, followed by a monotonic
decay. Parameter estimates use p̂ = 1 / (x̄ + 1).

6. Negative Binomial Distribution

An extension of the Geometric distribution, capturing the total failures experienced before hitting exactly r target successes.

PMF: P(X = x) = x + r - 1Cr - 1 pr qx, x = 0, 1, 2, ...

Moments: E[X] = rq / p, Var(X) = rq / p2

HIGH-YIELD EXAM PROPERTIES

• Mean vs. Variance Constraint: The variance is strictly larger than the mean (rq/p2 > rq/p since 0 < p < 1),
structurally representing the phenomenon of overdispersion.

APPLIED FIELD DIAGNOSTICS & IMPLEMENTATION

• Real-World Context: Modeling consumer purchasing behavior or tracking infectious disease case distributions
where clusters cause variance to expand.
• Data Diagnostics: Calculated sample metrics reveal s2 > x̄. Method of Moments yields parameters: p̂ = x̄ / s2 and r̂
= x̄2 / (s2 - x̄).

7. Hypergeometric Distribution

Governs sampling from a finite population completely without replacement. Because drawn units are not replaced, the
success probability shifts on each draw.

PMF: P(X = x) = [ MCx · N - MCn - x ] / NCn

Parameters: N (Population size), M (Total successes in population), n (Sample size drawn)

UGC NET English - Statistics (Code 107) Page 3 of 7

 APPLIED FIELD DIAGNOSTICS & IMPLEMENTATION

• Prerequisites: Finite total population, sampling without replacement, resulting in dependent trials.
• Data Diagnostics: Variance equations incorporate the Finite Population Correction (FPC) factor: (N - n) / (N - 1).
When population size N → ∞, it converges to a standard Binomial model.

UGC NET English - Statistics (Code 107) Page 4 of 7

 PART 2: STANDARD CONTINUOUS UNIVARIATE DISTRIBUTIONS

Continuous distributions dictate random variables capable of taking any infinite value across real intervals, depending heavily
on calculus integrations.

1. Continuous Uniform Distribution — U(a, b)

Also termed the rectangular distribution, it enforces an identical probability density across every sub-interval of equal length
within its bounded range.

PDF: f(x) = 1 / (b - a), a≤x≤b

Moments: E[X] = (a + b) / 2, Var(X) = (b - a)2 / 12
MGF: MX(t) = [ ebt - eat ] / [ t(b - a) ]

APPLIED FIELD DIAGNOSTICS & IMPLEMENTATION

• Real-World Context: Modeling random waiting times for a scheduled transit train that arrives precisely at fixed 20-
minute intervals.
• Estimation Profile: Maximum Likelihood Estimators (MLE) are defined entirely by the sample order extremes: â =
X(1) (sample minimum) and b̂ = X(n) (sample maximum).

2. Normal (Gaussian) Distribution — N(μ, σ2)

The core structural baseline of parametric statistics. Its characteristic bell curve profile naturally emerges whenever multiple
small independent variations are combined.

2
PDF: f(x) = [ 1 / (σ √(2π)) ] · e-0.5 · [ (x - μ) / σ ] , -∞ < x < ∞
22
MGF: MX(t) = eμt + 0.5 · σ t

HIGH-YIELD EXAM PROPERTIES

• Symmetry and Moments: Symmetric about μ, forcing all odd central moments to zero (μ3 = μ5 = ... = 0).
Skewness β1 = 0, Kurtosis β2 = 3.

• Points of Inflexion: The PDF curve shifts concavity precisely at coordinates x = μ ± σ.

• Linear Stability: Any linear combination of independent Normal variables remains perfectly Normally distributed.

APPLIED FIELD DIAGNOSTICS & IMPLEMENTATION

• The Central Limit Theorem (CLT): Explains why natural measurements (heights, error distributions) form bell
curves: the sum of many small independent effects converges to normality as sample size increases (n ≥ 30).
• Data Diagnostics: Verify if Mean = Median = Mode. Check data spread against the Empirical Rule: 68.27%
within ±1σ, 95.45% within ±2σ, and 99.73% within ±3σ. Standard ratios are fixed at QD : MD : SD → 10 : 12 : 15.

3. Exponential Distribution — Exp(λ)

Models the continuous time or distance that elapses between sequential event arrivals generated by a stable, memoryless
Poisson process.

UGC NET English - Statistics (Code 107) Page 5 of 7

 PDF: f(x) = λ e-λx, x ≥ 0, λ>0
Moments: E[X] = 1 / λ, Var(X) = 1 / λ2
MGF: MX(t) = λ / (λ - t) = (1 - t/λ)-1 for t < λ

HIGH-YIELD EXAM PROPERTIES

• Continuous Memoryless Property: The unique continuous distribution with this property: P(X > s + t | X > s) =
P(X > t). The probability of future survival remains independent of the current age or elapsed time.

APPLIED FIELD DIAGNOSTICS & IMPLEMENTATION

• Real-World Context: Baseline model for reliability engineering and survival analysis, tracking the operational
lifespan of electronic components before failure.
• Data Diagnostics: Check if the sample mean is roughly equal to the sample standard deviation (x̄ &inapprox; s).
Parameter estimation yields λ̂ = 1 / x̄.

4. Gamma Distribution

Models the total continuous waiting time required for exactly α independent Poisson arrivals to occur.

PDF: f(x) = [ λα / Γ(α) ] · xα - 1 e-λx, x≥0

Moments: E[X] = α / λ, Var(X) = α / λ2
MGF: MX(t) = (1 - t/λ)-α

HIGH-YIELD EXAM PROPERTIES

• Special Cases: Setting shape parameter α = 1 yields the Exponential distribution. Setting α = n/2 and λ = 1/2
matches a Chi-Square (χ2) distribution with n degrees of freedom.
• Additive Convolutions: If variables are independent and share a scale parameter λ, their shape parameters add
directly: (X + Y) ∼ Gamma(α1 + α2, λ).

APPLIED FIELD DIAGNOSTICS & IMPLEMENTATION

• Real-World Context: Widely applied in meteorology to model precipitation volumes, or in queuing theory to track
multi-stage service times.
• Estimation Profile: Method of Moments estimates follow via: λ̂ = x̄ / s2 and α̂ = x̄2 / s2.

5. Beta Distributions (First and Second Kind)

Highly flexible distributions commonly applied in Bayesian analytics to establish prior distributions for bounded parameters.

Beta First Kind [BetaI(α, β)]: f(x) = [ 1 / B(α, β) ] · xα - 1 (1 - x)β - 1, 0≤x≤1

Beta Second Kind [BetaII(α, β)]: f(x) = [ 1 / B(α, β) ] · [ xα - 1 / (1 + x)α + β ], x≥0

UGC NET English - Statistics (Code 107) Page 6 of 7

6. Cauchy Distribution

A symmetric, bell-shaped distribution that serves as a notable counter-example in statistical theory due to its extremely heavy
tails.

PDF: f(x) = 1 / [ π · (1 + (x - x0)2) ], -∞ < x < ∞

HIGH-YIELD EXAM PROPERTIES

• Non-Existence of Moments: The tails drop off slowly, causing the required expectation integrals to fail absolute
convergence tests. The Mean, Variance, and MGF do not exist for the Cauchy distribution.
• Reproductive Paradox: If X1, ..., Xn are independent standard Cauchy variables, their sample average X̄ follows
the exact same standard Cauchy distribution. Increasing the sample size fails to reduce variance, violating the
standard Law of Large Numbers.

APPLIED FIELD DIAGNOSTICS & IMPLEMENTATION

• Handling Extreme Outliers: Because the theoretical mean and variance are nonexistent, standard sample means
(x̄) and variances (s2) fail to stabilize. Analysts must use robust alternatives like the **Sample Median** and
**Interquartile Range (IQR)** to summarize the data.

PART 3: QUICK REFERENCE SUMMARY MATRIX

DISTRIBUTION TYPE MEAN VS VARIANCE UNIQUE SIGNATURE PROPERTY

Binomial Discrete Mean > Variance Tracks independent success counts across fixed trials;
reproductive.

Poisson Discrete Mean = Variance Governs rare event counts; standard limiting form of the
Binomial.

Negative Binomial Discrete Variance > Mean Models count data exhibiting structural overdispersion.

Geometric Discrete Variance > Mean The only discrete distribution possessing the memoryless
trait.

Normal Continuous Independent Parameters Symmetric; mesokurtic profile; stable under linear
transformations.

Exponential Continuous Mean = SD The only continuous distribution possessing the

memoryless trait.

Cauchy Continuous Do Not Exist Heavy-tailed; sample averages fail to compress baseline
variance.

UGC NET English - Statistics (Code 107) Page 7 of 7