Deriving Mean and Variance of Continuous Probability Distribution

The moment generating function (MGF) for a continuous random variable X is defined as M_X(t) = E[e^(tX)] = ∫_{-∞}^∞ e^(tx)f(x)dx, where f(x) is the probability density function. For an exponential distribution with f(x) = ke^(-kx) for x ≥ 0, the MGF evaluates to M_X(t) = k/(k-t) for t < k. The r-th moment about the origin μ'_r is obtained by taking the r-th derivative of the MGF with respect to t and evaluating at t=0. Alternatively, for distributions whose MGF can be expressed as a power series, μ'_r = r! × coefficient of t^r in the series expansion. For the exponential distribution with parameter k, the first four moments are μ'_1 = 1/k, μ'_2 = 2/k², μ'_3 = 6/k³, and μ'_4 = 24/k⁴, yielding variance σ² = μ'_2 - (μ'_1)² = 1/k².

The Moment Generating Function (MGF) for a continuous random variable X is defined as M_X(t) = E[e^(tX)] = ∫e^(tx)f(x)dx, where f(x) is the probability density function. The r-th moment about the origin, v_r = E[X^r], is obtained by taking the r-th derivative of the MGF with respect to t and evaluating at t=0. This provides a systematic method to generate all moments of the distribution.

The moment generating function (MGF) of a continuous random variable X is defined as M_X(t) = E[e^(tX)] = ∫ from -∞ to ∞ of e^(tx) f(x) dx. For a uniform distribution on [0,1] with f(x) = 1, the MGF is calculated as M_X(t) = ∫ from 0 to 1 of e^(tx) dx = (e^t - 1)/t. The first moment (mean) is found by taking the derivative of the MGF with respect to t and evaluating at t=0, yielding μ₁ = 1/2. This demonstrates how MGFs provide a systematic method for computing moments of continuous distributions through integration and differentiation.

The Moment Generating Function (MGF) is defined as M_X(t) = E[e^(tX)]. For continuous distributions, M_X(t) = ∫e^(tx)f(x)dx; for discrete, M_X(t) = Σe^(tx)P(X=x). The MGF uniquely determines the distribution and allows easy moment calculation. For exponential distribution f(x) = (1/θ)e^(-x/θ), M_X(t) = 1/(1 - θt). For normal distribution with mean μ and variance σ², M_X(t) = e^(μt + (σ²t²)/2). For standard normal, M_Z(t) = e^(t²/2). For Y = aX + b, M_Y(t) = e^((μa + b)t + (σ²a²t²)/2), showing linear transformations of normal variables remain normal. For Poisson with parameter λ, M_X(t) = e^(λ(e^t - 1)). For chi-squared with k degrees of freedom, M_X(t) = (1 - 2t)^(-k/2). The MGF method involves finding M_Y(t) = E[e^(tY)], recognizing the distribution, and reading off the PDF.

The moment generating function (MGF), denoted by M_X(t), generates moments of a random variable. For discrete variables: M_X(t) = Σ e^(tx)·P(X = x). For continuous variables: M_X(t) = ∫ e^(tx)·f(x) dx. Moments are obtained by differentiating the MGF and evaluating at t = 0: E[X^k] = M_X^(k)(0). The first derivative gives the mean, the second derivative gives the second moment. The MGF provides a powerful tool for analyzing distributions and deriving moments.

This example demonstrates calculating moments from continuous distribution data. Given class intervals: 0-10, 10-20, 20-30, 30-40, 40-50, 50-60 with frequencies: 1, 6, 10, 15, 11, 7. Class marks xi: 5, 15, 25, 35, 45, 55. Choose a = 35 as reference point. Calculate di: -30, -20, -10, 0, 10, 20. Calculate Σfi = 50, Σfidi = 0, Σfidi² = 820, Σfidi³ = -18,000, Σfidi⁴ = 41,000. Then μ1' = 0/50 = 0, μ2' = 820/50 = 16.4, μ3' = -18,000/50 = -360, μ4' = 41,000/50 = 820. Using relationships: μ1 = 0, μ2 = μ2' - (μ1')² = 16.4, μ3 = μ3' - 3μ2'μ1' + 2(μ1')³ = -360, μ4 = μ4' - 4μ3'μ1' + 6μ2'(μ1')² - 3(μ1')⁴ = 820. Then β1 = μ3² / (μ2)^(3/2) = 1945.1, and β2 = μ4 / (μ2)² = 3.05. Since β2 > 3, the distribution is leptokurtic.

Using the raw moments calculated for the continuous distribution f(x) = 2x on [0,1], we apply the central moment formulas. With μ̄₁ = 2/3, μ̄₂ = 1/2, μ̄₃ = 2/5, and μ̄₄ = 1/3, the central moments are: μ₂ = μ̄₂ - μ̄₁² = 1/2 - (4/9) = 1/18; μ₃ = μ̄₃ - 3μ̄₂μ̄₁ + 2μ̄₁³ = 2/5 - 3*(1/2)*(2/3) + 2*(8/27) = -1/135; μ₄ = μ̄₄ - 4μ̄₃μ̄₁ + 6μ̄₂μ̄₁² - 3μ̄₁⁴ = 1/3 - 4*(2/5)*(2/3) + 6*(1/2)*(4/9) - 3*(16/81) = 1/135. The negative μ₃ indicates negative skewness, and the coefficient calculations confirm this.

For continuous series, construct a table with columns for mid-values (m), assumed mean (a), deviations (d = (m - a)/h), frequency (f), fd, fd², and fd⁴. Choose a central value as assumed mean to minimize calculations. Calculate class interval size (h) as difference between consecutive boundaries. Compute raw moments: μ1' = (1/n) × Σ(fd) × h, μ2' = (1/n) × Σ(fd²) × h², μ3' = (1/n) × Σ(fd³) × h³, μ4' = (1/n) × Σ(fd⁴) × h⁴. Convert raw moments to central moments using formulas: μ2' = μ2'' - (μ1')², μ3' = μ3'' - 3μ2''μ1' + 2(μ1')³, μ4' = μ4'' - 4μ3''μ1' + 6μ2''(μ1')² - 3(μ1')⁴.

Skewness (third central moment) measures distribution lopsidedness; symmetric distributions have zero skewness. Left-skewed distributions have negative skewness, right-skewed have positive. For near-normal distributions, median and mode relate to mean and skewness. Kurtosis (fourth central moment) measures tail heaviness compared to normal distribution; normal distribution has kurtosis of 3. Heavy-tailed distributions have high kurtosis, light-tailed have low. Kurtosis must be ≥ γ² - 2, with equality for binary distributions. Higher moments beyond fourth-order are higher-order statistics involving non-linear data combinations, useful for describing additional shape parameters but harder to estimate requiring larger samples.

Moments are fundamental statistical measures describing probability distributions. Moments about origin use notation μ₀, μ₁, μ₂... while moments about the mean use μ₁', μ₂'... For continuous variables, moments are calculated via integrals; for discrete variables, via summations. The first moment about origin equals the mean. The first moment about the mean is always zero. The second moment about the mean equals the variance (σ²), representing data spread. The third moment about the mean (μ₃') measures skewness, calculated as E[(x - x̄)³], which simplifies to μ₃ using binomial expansion. The fourth moment about the mean (μ₄') measures kurtosis, calculated as E[(x - x̄)⁴], which simplifies to μ₄. These formulas enable systematic calculation of distribution shape parameters.

For continuous variables, the mean is E[X] = ∫ x f_X(x) dx and the variance is Var(X) = E[X²] - (E[X])² = ∫ x² f_X(x) dx - (∫ x f_X(x) dx)². For covariance: Cov(X,Y) = E[XY] - E[X]E[Y], where E[XY] = ∬ xy f(x,y) dx dy. The correlation coefficient is ρ = Cov(X,Y) / (σ_X × σ_Y), where σ_X = √Var(X) and σ_Y = √Var(Y). The correlation coefficient ranges from -1 to +1 and measures the strength and direction of the linear relationship. To calculate the conditional PDF f(x|y): (1) Identify the region where f(x,y) > 0 for a given y, (2) Calculate f_Y(y) = ∫ f(x,y) dx, (3) Then f(x|y) = f(x,y) / f_Y(y).

Joint probability distributions describe multiple random variables together. For discrete variables, P(X=x,Y=y) gives combined probabilities. Marginal distributions are found by summing over other variables: P(X=x) = Σᵧ P(X=x,Y=y). Conditional distributions P(X=x|Y=y) = P(X=x,Y=y)/P(Y=y). Independence means P(X=x,Y=y) = P(X=x)P(Y=y). For continuous variables, integrate joint PDFs over regions to find probabilities. Marginal PDFs are found by integrating over other variables: f_X(x) = ∫f(x,y)dy. Conditional PDFs f(x|y) = f(x,y)/f_Y(y). Expectation E[g(X,Y)] = ∫∫g(x,y)f(x,y)dx dy. E[X+Y] = E[X]+E[Y] always holds. Var(X+Y) = Var(X)+Var(Y)+2Cov(X,Y). Covariance Cov(X,Y) = E[XY] - E[X]E[Y] measures linear relationship. Correlation coefficient ρ = Cov(X,Y)/(σ_Xσ_Y) ranges from -1 to 1. Conditional expectation E[X|Y=y] = ∫x f(x|y)dx. Conditional variance Var(X|Y=y) = E[X²|Y=y] - (E[X|Y=y])².

To find marginal distributions from a joint table, sum the probabilities across rows (for X) or columns (for Y). The expected value of a random variable from a joint distribution is calculated as E[X] = Σ_x xP(X=x) for discrete, or ∫x f_X(x) dx for continuous. The expected value of a function g(X,Y) is E[g(X,Y)] = ΣΣ g(x,y)P(x,y) for discrete, or ∫∫ g(x,y)f(x,y)dx dy for continuous. The variance is Var(X) = E[X²] - (E[X])². The covariance is Cov(X,Y) = E[XY] - E[X]E[Y]. The correlation coefficient is ρ = Cov(X,Y) / (σ_Xσ_Y). The conditional expectation E[Y|X=x] is calculated as E[Y|X=x] = Σ_y yP(Y=y|X=x) for discrete, or ∫y f(y|x) dy for continuous. The conditional PDF is f(y|x) = f(x,y)/f_X(x). The conditional variance is Var(Y|X=x) = E[Y²|X=x] - (E[Y|X=x])². The conditional expectation of a function g(X,Y) given X=x is E[g(X,Y)|X=x] = Σ_y g(x,y)P(Y=y|X=x) for discrete, or ∫g(x,y)f(y|x) dy for continuous. For continuous random variables where one depends on another (e.g., Y depends on X), the joint PDF is defined over the region satisfying the dependency. The constant is found by integrating the function over this region and setting equal to 1.

This section covers covariance, correlation coefficient, and conditional expectation for continuous random variables. The instructor explains how to compute covariance by first finding expected values, then computing the expected value of the product of deviations. The correlation coefficient is introduced as a normalized measure of linear relationship, bounded between -1 and 1. The section demonstrates computing conditional expectation by integrating the conditional PDF over the variable of interest. The instructor shows how to derive the conditional expectation for a specific joint PDF, emphasizing the importance of understanding the conditional distribution before computing expectations.

Covariance measures the relationship between two random variables: Cov(X,Y) = E[XY] - μXμY. The correlation coefficient ρ = Cov(X,Y)/(σXσY) ranges from -1 to +1, indicating strength and direction of linear relationship. For continuous variables, expected values are found by integrating x·f(x) over the range, and variance is Var(X) = E[X²] - (E[X])². For discrete variables, joint probability distributions are represented in tables. Marginal distributions are obtained by summing joint probabilities across rows or columns. Expected values are calculated as E[X] = Σx·P(X=x) and E[XY] = ΣΣxy·P(X=x,Y=y). Correlation interpretation: |ρ| < 0.5 indicates weak correlation, 0.5-0.8 moderate, and ≥0.8 strong.

The Chebyshev Inequality states that for any random variable with finite mean μ and variance σ², the probability that the random variable deviates from its mean by at least c units is at most σ²/c²; this inequality demonstrates that smaller variance implies tighter concentration around the mean, and when measuring deviation in terms of standard deviations (kσ), the probability of being k or more standard deviations away from the mean is at most 1/k², making it a powerful tool for bounding probabilities without knowing the exact distribution.

Chebyshev's Inequality states that for any random variable with finite mean μ and variance σ², P(|X - μ| ≥ kσ) ≤ 1/k². Alternative forms include P(|X - μ| ≥ c) ≤ σ²/c² and P(|X - μ| < kσ) ≥ 1 - 1/k². This inequality applies to any distribution (positive or negative values) with finite mean and variance. For binomial distribution (n=10, p=0.5), P(|X - 5| ≥ 2σ) ≈ 0.021 < 0.25. For geometric distribution (p=0.25), P(|X - μ| ≥ 2σ) ≈ 0.056 < 0.25. These examples demonstrate Chebyshev's Inequality provides universal bounds regardless of distribution shape. Markov's Inequality applies only to non-negative variables, while Chebyshev's applies to any variable with finite mean and variance.

This video demonstrates two fundamental statistical concepts: Chebyshev's Inequality, which provides an upper bound for the probability that a random variable deviates from its mean by more than k standard deviations (P(|X-μ| ≥ kσ) ≤ 1/k²), and the Central Limit Theorem, which states that the sum of independent, identically distributed random variables approaches a normal distribution as the sample size increases. The video applies these concepts to solve problems involving dice sums and sample means from normal populations.

This section introduces Chebyshev's Inequality as a universal bound applicable to any random variable with finite mean and variance. The inequality states that P(|X - μ| ≥ kσ) ≤ 1/k², meaning the probability of deviating by k standard deviations or more is at most 1/k². The lecture explains that this result is derived by applying Markov's Inequality to the squared deviation (X - μ)², which is always non-negative. The complementary form P(|X - μ| < kσ) ≥ 1 - 1/k² shows that most probability mass concentrates near the mean. Examples demonstrate that while Chebyshev's bounds are mathematically guaranteed, actual distributions often satisfy much tighter bounds.

For independent Bernoulli variables X₁, ..., Xₙ with parameter P, the sample mean X̄ = (1/n)ΣXᵢ has variance Var(X̄) = P(1-P)/n. This follows from: each Xᵢ has variance P(1-P), independence implies Var(ΣXᵢ) = nP(1-P), and Var(aX) = a²Var(X). Chebyshev's inequality states P(|X̄ - P| ≥ ε) ≤ Var(X̄)/ε². Applying this with ε = √5/n gives P(|X̄ - P| ≥ √5/n) ≤ P(1-P)n/5.