Paper: arXiv 2408.07405

Authors: Ajay Jasra, Mohamed Maama, Aleksandar Mijatović

Abstract

In this paper we consider the modeling of measurement error for fund returns data. In particular, given access to a time-series of discretely observed log-returns and the associated maximum over the observation period, we develop a stochastic model which models the true log-returns and maximum via a Lévy process and the data as a measurement error there-of. The main technical difficulty of trying to infer this model, for instance Bayesian parameter estimation, is that the joint transition density of the return and maximum is seldom known, nor can it be simulated exactly. Based upon the novel stick breaking representation of [“12”] we provide an approximation of the model. We develop a Markov chain Monte Carlo (MCMC) algorithm to sample from the Bayesian posterior of the approximated posterior and then extend this to a multilevel MCMC method which can reduce the computational cost to approximate posterior expectations, relative to ordinary MCMC. We implement our methodology on several applications including for real data.

Complexity vs Empirical Score

  • Math Complexity: 9.0/10
  • Empirical Rigor: 7.0/10
  • Quadrant: Holy Grail — high math complexity, high empirical rigor

Why this score: The paper employs advanced mathematics including Lévy processes, stick-breaking representations, and multilevel Monte Carlo theory. It demonstrates empirical rigor through a developed MCMC algorithm, implementation on real fund data, and analysis of computational cost reduction.

Research Flowchart

  flowchart TD
  A["Research Goal<br>Model measurement error<br>in discretely observed<br>financial returns and maxima"] --> B["Methodology<br>Develop stochastic model<br>using Lévy process &<br>stick-breaking approximation"]
  B --> C["Data Inputs<br>Time-series of log-returns<br>and associated maxima"]
  B --> D["Computational Process<br>Bayesian inference via<br>Multilevel MCMC<br>for posterior sampling"]
  C --> D
  D --> E["Key Findings<br>Approximated posterior distributions<br>for fund returns<br>and improved computational efficiency"]

Evidence Card

What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card

Proposes

The paper proposes a new Bayesian model for discretely observed fund returns with associated maxima, using a latent Lévy process, and introduces a multilevel MCMC method to reduce computational cost.

Source sentence
We develop a Markov chain Monte Carlo (MCMC) algorithm to sample from the Bayesian posterior of the approximated posterior and then extend this to a multilevel MCMC method which can reduce the computational cost to approximate posterior expectations, relative to ordinary MCMC.

Assumes 5

  • Data are observed over unit time intervals. data
    Source sentence
    We consider the regular in time observation of the log-returns and the maximal such value over several time instances. We assume data observed over unit time intervals, but this is simply for notational
  • The Lévy process has a unique maximum time τt almost surely. model
    Source sentence
    For all Lévy processes considered in this paper, τ t is almost surely unique in the interval (0, t).
  • The stick‑breaking approximation error decays geometrically with an accuracy parameter. model
    Source sentence
    ... provide a means to simulate the value of the process and it’s maximum subject to an arbitrarily small error which decays geometrically in an accuracy parameter.
  • Markov chains used in the analysis are started in stationarity, reversible and uniformly ergodic. statistical
    Source sentence
    All the Markov chains are started in stationarity are reversible with respect to their invariant measure and uniformly ergodic.
Show 1 more
  • Gaussian errors are not necessary; any tractable density could be used. statistical
    Source sentence
    We remark that Gaussian errors are not necessary and indeed any (tractable) probability density could be used if needed.

Data

  • Simulated BMG model data synthetic unknown
    Source sentence
    In the BMG model, the parameters to be estimated are b and σ, which are assigned independent Gamma priors. A data set with T observations is simulated with b = 1 and σ = 0.5.
  • S&P 500 weekly log‑returns · Yahoo Finance market-prices public
    January 1, 2021 to November 1, 2023 · S&P 500 index
    Source sentence
    For the real data, consisting of weekly log‑returns from the S&P 500 index, obtained from Yahoo Finance, covering the period from January 1, 2021, to November 1, 2023, the priors were Ga(0.1, 0.1) for both parameters.
  • Weekly hedge fund return data fundamentals unknown
    Source sentence
    the observation data to be assimilated are finance data, specifically weekly hedge return fund data.

Recognised sources: Yahoo Finance

Costs

Not applicable The paper reports no trading or execution result.

Finds 2

  • The multilevel Monte Carlo method reduces the computational cost to achieve a given mean square error compared with standard MCMC. theoretical
    Source sentence
    we show how to utilize the multilevel Monte Carlo (MLMC) method to improve the cost to achieve a given mean square error versus the afore-mentioned MCMC approach.
  • The stick‑breaking approximation provides an arbitrarily small error that decays geometrically, enabling faithful Bayesian estimation. theoretical
    Source sentence
    This latter property is key in producing Bayesian estimation that is as faithful as possible to the original model that we develop.

Checked by others

No other paper in the archive has been found to support or contradict these findings (compared with 11 related papers). Coverage is partial, so this is neither replication nor refutation.

Code

No repository by the authors found in the paper.

Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.