Understanding the Gamma Distribution: Applications, Theory, and Its Link to the Exponential

Introduction

The lesson expands on the introductory material about the gamma distribution, explaining why it is useful in actuarial science and other fields. It covers practical applications, the gamma function, derivation of the standard and two‑parameter gamma PDFs, moment‑generating function, mean, variance, and the relationship to the exponential distribution.

Real‑World Applications

• Insurance claim severity – Models the total amount of loss per claim; data are highly skewed, making the gamma distribution a natural fit.
• Biology – Describes time variability in cell proliferation and gene expression.
• Psychology – Used for modeling scholastic outcomes. These examples illustrate why actuaries and researchers prefer the gamma distribution over simpler models when data exhibit positive skewness.

The Gamma Function

• Defined as (\Gamma(\alpha)=\int_0^{\infty} y^{\alpha-1}e^{-y}\,dy).
• Extends the factorial: (\Gamma(n+1)=n!) for non‑negative integers.
• Key properties:
• (\Gamma(\alpha)= (\alpha-1)\Gamma(\alpha-1)) (proved by integration by parts).
• (\Gamma(1/2)=\sqrt{\pi}).
• Domain: all real numbers except zero and negative integers; positive for (\alpha>0).

Deriving the Standard Gamma PDF

Starting from the gamma function and dividing by (\Gamma(\alpha)) yields a valid probability density function (PDF) for (\alpha>0) and (y>0): [ f_Y(y)=\frac{1}{\Gamma(\alpha)}y^{\alpha-1}e^{-y} ] This is the standard gamma distribution (scale parameter (\theta=1)).

Extending to the Two‑Parameter Gamma Distribution

If (X=\theta Y) with (\theta>0), the transformation gives: [ f_X(x)=\frac{1}{\Gamma(\alpha)\theta^{\alpha}}x^{\alpha-1}e^{-x/\theta},\quad x>0 ] An alternative parameterization uses a rate parameter (\beta=1/\theta): [ f_X(x)=\frac{\beta^{\alpha}}{\Gamma(\alpha)}x^{\alpha-1}e^{-\beta x},\quad x>0 ] Both forms are common in textbooks.

Moment‑Generating Function (MGF)

For (X\sim\text{Gamma}(\alpha,\theta)): [ M_X(t)=E[e^{tX}] = \frac{1}{(1-\theta t)^{\alpha}},\qquad t<\frac{1}{\theta} ] The MGF exists only for (t) below the reciprocal of the scale parameter.

Mean and Variance

• Mean: (E[X]=\alpha\theta) (first derivative of the MGF at (t=0)).
• Second moment: (E[X^2]=\alpha(\alpha+1)\theta^2).
• Variance: (\operatorname{Var}(X)=E[X^2]-E[X]^2 = \alpha\theta^2). These formulas simplify calculations in actuarial models.

Connection to the Exponential Distribution

Setting (\alpha=1) reduces the gamma PDF to: [ f_X(x)=\frac{1}{\theta}e^{-x/\theta},\quad x>0 ] which is exactly the exponential distribution. Consequently, the exponential is a special case of the gamma, sharing the same MGF (with (\alpha=1)), mean ((\theta)), and variance ((\theta^2)).

Summary of Key Formulas

• PDF (shape (\alpha), scale (\theta)): (f(x)=\frac{1}{\Gamma(\alpha)\theta^{\alpha}}x^{\alpha-1}e^{-x/\theta}).
• PDF (shape (\alpha), rate (\beta)): (f(x)=\frac{\beta^{\alpha}}{\Gamma(\alpha)}x^{\alpha-1}e^{-\beta x}).
• MGF: (M(t)= (1-\theta t)^{-\alpha}), (t<1/\theta).
• Mean: (\alpha\theta).
• Variance: (\alpha\theta^2).

Looking Ahead

The next lesson will cover joint distributions of random variables, building on the concepts introduced here.

The gamma distribution is a versatile tool for modeling positively skewed data, especially claim severity in insurance, and it seamlessly generalizes to the exponential distribution, making it essential knowledge for actuaries and statisticians.