On the characteristic function of the asymmetric Student's $t$-distribution and an integral involving the sine function

This paper derives a new closed-form formula for the characteristic function of the asymmetric Student's $t$-distribution by establishing a novel integral representation involving the sine function and the exponential integral, which subsequently yields a closed-form expression for a specific limit involving modified Bessel and Struve functions.

The Gist

Imagine you are a statistician trying to understand the "personality" of a specific type of data. In the world of finance and risk, data doesn't always behave politely. Sometimes it has a long tail on the left (lots of small losses) and a short tail on the right (rare, massive gains), or vice versa. This is called skewness.

For decades, statisticians have used a famous tool called the Student's t-distribution to model this kind of "wild" data. However, the classic version is perfectly symmetrical (like a bell curve that's been squashed). To handle real-world chaos, researchers invented the Asymmetric Student's t-distribution (AST). It's like the classic tool, but with a flexible handle that lets you tilt it left or right to fit the data perfectly.

The Problem:
While we knew how to describe the shape of this new tool (its Probability Density Function), we were missing its "ID card." In statistics, this ID card is called the Characteristic Function (CF). It's a mathematical magic wand that allows you to predict how the data will behave when you combine it with other data, calculate risks, or simulate future scenarios. Without the CF, the AST distribution is like a car with a great engine but no steering wheel—you can't really drive it effectively.

Previous attempts to write down this "steering wheel" (the formula) were messy, full of errors, or so complicated they were useless for practical math.

The Solution:
Robert Gaunt, the author of this paper, has built a brand new, cleaner, and more reliable steering wheel for the AST distribution.

Here is how he did it, using some simple analogies:

1. The "Sine Wave" Puzzle

To build the steering wheel, Gaunt first had to solve a tricky math puzzle involving a sine wave (a smooth, wavy line) and a specific type of curve.

• The Analogy: Imagine you are trying to measure the total "wobble" of a sine wave as it passes through a series of increasingly complex filters (mathematical barriers).
• The Gap: Mathematicians knew how to measure this wobble for most filters, but they hit a wall when the filters were whole numbers (like 2, 3, 4...). It was like having a map for a city that suddenly stops at a specific street, leaving a "No Entry" sign for the rest.
• The Breakthrough: Gaunt derived a new, closed-form formula (a neat, finished equation) to measure this wobble for any whole number filter. He didn't just guess; he used a technique called "partial fraction decomposition," which is like taking a complex Lego structure apart into its individual bricks, solving the problem for each brick, and snapping them back together.

2. The "Limit" Trick

In his process, he also solved a mystery about what happens when a mathematical function gets infinitely close to a specific number (a "limit").

• The Analogy: Imagine you are walking toward a cliff. You know exactly where you are at every step, but the cliff edge itself is undefined. Gaunt figured out exactly what the "edge" looks like, providing a smooth landing spot where others saw a drop-off. This helps mathematicians handle cases where the standard formulas break down.

3. The New "Steering Wheel" (The Characteristic Function)

With the sine wave puzzle solved, Gaunt assembled the final formula for the AST distribution.

• Why it's better: The old formulas were like a tangled ball of yarn involving many different types of special functions (hypergeometric functions, etc.). Gaunt's new formula is streamlined. It uses a specific set of "special functions" (Modified Bessel and Struve functions) that are well-understood and easier for computers to calculate.
• The Safety Net: He noticed that for certain specific settings (when the "degrees of freedom" are odd numbers like 1, 3, 5), the standard formula would divide by zero and crash. So, he created a special "emergency backup" formula for those specific cases, ensuring the tool never breaks.

The Payoff

Why does this matter to a regular person?

1. Better Risk Management: Financial institutions use these distributions to model stock market crashes or insurance claims. A better formula means more accurate predictions of risk, potentially saving money and preventing disasters.
2. Simpler Math: By providing a clean, closed-form formula, Gaunt makes it much easier for software to simulate these complex scenarios quickly.
3. Connecting the Dots: His work proves that this new, complex tool (AST) naturally simplifies back into the classic Student's t-distribution when you set the "skewness" to zero. It's like showing that a high-tech sports car is still fundamentally a car, just with more options.

In Summary:
Robert Gaunt took a complex, slightly broken mathematical tool used to model real-world chaos, fixed the missing parts by solving a difficult integral puzzle, and handed back a polished, reliable, and easy-to-use version. He didn't just patch the hole; he rebuilt the foundation so that future mathematicians and data scientists can drive their models safely and efficiently.

Technical Summary

Here is a detailed technical summary of the paper "On the characteristic function of the asymmetric Student's t-distribution and an integral involving the sine function" by Robert E. Gaunt.

1. Problem Statement

The paper addresses two primary mathematical gaps:

1. The Characteristic Function (CF) of the Asymmetric Student's t-distribution (AST): While the AST distribution (introduced by [12]) is a widely used generalization of the skewed Student's t-distribution for modeling financial data, a correct, closed-form formula for its characteristic function was missing. Previous attempts (specifically by [7]) provided formulas involving generalized hypergeometric functions and other special functions, but these were identified as erroneous due to singularities at integer degrees of freedom ($\nu = 1, 3, 5, \dots$).
2. Evaluation of a Specific Definite Integral: The derivation of the AST's CF requires evaluating the integral $\int_0^\infty \frac{\sin(ax)}{(b^2 + x^2)^n} dx$ for positive integers $n$. While formulas exist for non-integer exponents or for the specific case $n=1$, no closed-form solutions were available in the literature for $n = 2, 3, 4, \dots$.

2. Methodology

The author employs a combination of complex analysis, partial fraction decomposition, and properties of special functions to solve these problems.

• Integral Evaluation (The Core Tool):

  • The author derives a closed-form expression for $I_n = \int_0^\infty \frac{\sin(ax)}{(b^2 + x^2)^n} dx$ for $n \in \mathbb{Z}^+$.
  • Technique: The rational function $\frac{1}{(1+x^2)^n}$ is decomposed into partial fractions over the complex numbers. This allows the integral to be split into terms involving $\frac{e^{\pm iax}}{(x \pm i)^k}$.
  • Integration: The author utilizes a known integral formula involving the exponential integral function $Ei(x)$ derived via repeated integration by parts.
  • Simplification: By combining these terms and simplifying complex coefficients, a general summation formula is obtained.

• Derivation of the Limit:

  • Using the new integral formula, the author deduces a closed-form limit for the expression $\lim_{\nu \to n} \frac{I_{\nu-1/2}(x) - L_{1/2-\nu}(x)}{\sin(\pi\nu)}$, where $I$ is the modified Bessel function and $L$ is the modified Struve function. This resolves the singularity that occurs when $\nu$ is an integer in the standard formula (1.2).

• Derivation of the Characteristic Function:

  • The AST probability density function (PDF) is split into negative ($x \leq 0$) and positive ($x > 0$) components, each resembling a scaled Student's t-distribution.
  • The CF, $\phi_X(t) = E[e^{itX}]$, is split into real (cosine) and imaginary (sine) parts.
  • The Real Part is evaluated using Basset's integral formula (involving the modified Bessel function of the second kind, $K_\nu$).
  • The Imaginary Part is evaluated using the newly derived integral formula (involving $Ei$, $I_\nu$, and $L_\nu$).
  • Special care is taken to handle cases where the degrees of freedom $\nu$ are odd integers, where the standard Struve/Bessel combination becomes undefined.

3. Key Contributions and Results

A. New Integral Formula (Theorem 2.1)

The paper provides a closed-form solution for $\int_0^\infty \frac{\sin(ax)}{(b^2 + x^2)^n} dx$ for all $n \in \mathbb{Z}^+$. The result is expressed in terms of the exponential integral function $Ei(x)$ and finite sums.

• Significance: This fills a gap in the literature, as previous formulas were limited to non-integer exponents or $n=1$. The formula is explicitly given as:
  $$\int_0^\infty \frac{\sin(ax)}{(b^2 + x^2)^n} dx = \frac{1}{b^{2n-1}} \sum_{k=1}^n \binom{2n-k-1}{n-1} \frac{2^{k-2n}}{(k-1)!} \left[ \dots \right]$$
  (See Eq. 2.5 in the paper for the full expansion involving $Ei$).

B. Limit of Special Functions (Corollary 2.3)

A closed-form formula is derived for the limit $\lim_{\nu \to n} \frac{I_{\nu-1/2}(x) - L_{1/2-\nu}(x)}{\sin(\pi\nu)}$ for integer $n$.

• Significance: This resolves the indeterminacy in the standard integral representation (1.2) when the parameter $\rho$ is an integer, providing a bridge between the continuous and discrete parameter cases.

C. Characteristic Function of the AST Distribution (Theorem 3.1)

The paper presents the first correct, closed-form formula for the CF of the AST distribution.

• Structure: The CF is expressed as $\phi_X(t) = A + iB$, where:
  • The real part involves the modified Bessel function of the second kind ($K_\nu$).
  • The imaginary part involves the modified Bessel function of the first kind ($I_\nu$), the modified Struve function ($L_\nu$), and the exponential integral ($Ei$).
• Handling Singularities: The formula is split into two cases:
  1. $\nu \notin \{1, 3, 5, \dots\}$: Uses the standard Struve/Bessel combination.
  2. $\nu \in \{1, 3, 5, \dots\}$: Uses the new summation formula derived from the integral evaluation to avoid division by zero (which occurs due to $\cos(\pi\nu/2)$ in the denominator of the standard form).
• Simplicity: The author notes that this formula is simpler than previous (incorrect) attempts, avoiding generalized hypergeometric functions.

D. Generalization with Location and Scale (Corollary 3.4)

The results are extended to the AST distribution with location ($\mu$) and scale ($\sigma$) parameters, providing a complete CF for the most general form of the distribution used in practice.

4. Significance and Impact

• Correction of Literature: The paper corrects erroneous formulas previously published in [7], which contained singularities that made them unusable for integer degrees of freedom.
• Theoretical Completeness: It provides a theoretically satisfying and complete characterization of the AST distribution's CF, which is fundamental for statistical inference, simulation, and Fourier-based methods.
• Practical Application: Since the AST distribution is a popular alternative to the classical Student's t-distribution for modeling financial data (which often exhibits skewness and heavy tails), having an accurate CF allows for better model fitting, option pricing, and risk management.
• Mathematical Utility: The derived integral formulas and limits for special functions are valuable tools for researchers working in probability theory, mathematical physics, and applied mathematics involving Bessel and Struve functions.

In summary, Robert E. Gaunt successfully derives the missing closed-form characteristic function for the asymmetric Student's t-distribution by first solving a challenging class of definite integrals involving the sine function and the exponential integral, thereby resolving singularities in existing special function representations.