Truncated sequential guaranteed estimation for the Cox-Ingersoll-Ross models

Mohamed BEN ALAYA a , Thi-Bao Trâm NGÔ b and Serguei PERGAMENCHTCHIKOV a · 2025

Read Truncated sequential guaranteed estimation for the Cox-Ingersoll-Ross models by Mohamed BEN ALAYA a , Thi-Bao Trâm NGÔ b and Serguei PERGAMENCHTCHIKOV a (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2504-04923.txt.

Introduction

Mohamed Ben Alaya and co-authors design guaranteed estimators for an interest-rate model in a 2025 paper. The Cox-Ingersoll-Ross process models rates that wander but pull back toward a level. Estimating its drift from a limited watch is the task. Their truncated sequential method samples until it can promise a set accuracy, then stops, rather than always using the full watch.

For both one- and two-parameter versions, the mean-square error guarantee holds at any sample size, not just in the limit. As the allowed time grows, the procedures prove optimal among all methods using less average time. They also reach maximum-likelihood quality in far less observing time. Shorter data collection with the same trust is the prize.

Catalog record

CREATORMohamed BEN ALAYA a , Thi-Bao Trâm NGÔ b and Serguei PERGAMENCHTCHIKOV a
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE342 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2504-04923

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Truncated sequential guaranteed estimation for the Cox-Ingersoll-Ross models Mohamed BEN ALAYA a , Thi-Bao Trâm NGÔ b and Serguei PERGAMENCHTCHIKOV a Abstract The drift sequential parameter estimation problems for the Cox-Ingersoll-Ross (CIR) processes under the limited duration of observation are studied. Truncated sequential estimation methods for both scalar and two-dimensional parameter cases are proposed. In the non-asymptotic setting, for the proposed truncated estimators, the properties of guaranteed mean-square estimation accuracy are established. In the asymptotic formulation, when the observation time tends to infinity, it is shown that the proposed sequential procedures are asymptotically optimal among all possible sequential and non-sequential estimates with an average estimation time less than the fixed observation duration. It also turned out that asymptotically, without degrading the estimation quality, they significantly reduce the observation duration compared to classical non-sequential maximum likelihood estimations based on a fixed observation duration. keywords Cox-Ingersoll-Ross processes; guaranteed estimation method; truncated sequential estimation; parameter estimation; minimax estimation. 2010 MSC : Primary 44A10; 60F05; Secondary 62F12; 65C05. † † articletype: ARTICLE † † affiliation: a Laboratoire de Mathématiques Raphaël Salem, UMR 6085 CNRS-Université de Rouen Normandie, France; b Laboratoire de Mathématiques et Modélisation d’Évry, CNRS, UMR 8071, Université Évry Paris-Saclay, France 1 Introduction 1.1 Motivations In this paper, based on the sequential analysis approach, we develop a new truncated estimation method based on observations within the time interval [0,T] of the Cox-Ingersoll-Ross (CIR) process defined through the following stochastic differential equation \mathrm{d}X_{\mathchoice{t}{t}{\lower 0.75346pt\hbox{$\scriptstyle t$}}{\lower 0.75346pt\hbox{$\scriptscriptstyle t$}}}=(a-bX_{\mathchoice{t}{t}{\lower 0.75346pt\hbox{$\scriptstyle t$}}{\lower 0.75346pt\hbox{$\scriptscriptstyle t$}}})\mathrm{d}t+\sqrt{\sigma X_{\mathchoice{t}{t}{\lower 0.75346pt\hbox{$\scriptstyle t$}}{\lower 0.75346pt\hbox{$\scriptscriptstyle t$}}}}\mathrm{d}W_{\mathchoice{t}{t}{\lower 0.75346pt\hbox{$\scriptstyle t$}}{\lower 0.75346pt\hbox{$\scriptscriptstyle t$}}},\quad X_{\mathchoice{0}{0}{\lower 0.75346pt\hbox{$\scriptstyle 0$}}{\lower 0.75346pt\hbox{$\scriptscriptstyle 0$}}}=x>0,\quad 0\leq t\leq T\,, (1) where a>0 , b>0 and \sigma>0 are fixed parameters and (W_{\mathchoice{t}{t}{\lower 0.75346pt\hbox{$\scriptstyle t$}}{\lower 0.75346pt\hbox{$\scriptscriptstyle t$}}})_{\mathchoice{t\geq 0}{t\geq 0}{\lower 0.75346pt\hbox{$\scriptstyle t\geq 0$}}{\lower 0.75346pt\hbox{$\scriptscriptstyle t\geq 0$}}} is a standard Brownian motion. Similarly to Ben Alaya, Ngô and Pergamenchtchikov (2025) , we consider the sequential estimation problem for the parameters a and b under the condition that the diffusion coefficient \sigma is known. It should be noted that in this case the process ( 1 ) is ergodic (see, for example, in Ben Alaya and Kebaier (2010) , for the details) and has the ergodic density which defined as {\bf q}_{\mathchoice{a,b}{a,b}{\lower 0.75346pt\hbox{$\scriptstyle a,b$}}{\lower 0.75346pt\hbox{$\scriptscriptstyle a,b$}}}(z)=\frac{\beta^{\alpha}}{\Gamma(\alpha)}\,z^{\alpha-1}\,e^{-\beta z}\,{\bf 1}_{\mathchoice{\{z\geq 0\}}{\{z\geq 0\}}{\lower 0.75346pt\hbox{$\scriptstyle\{z\geq 0\}$}}{\lower 0.75346pt\hbox{$\scriptscriptstyle\{z\geq 0\}$}}}\,, (2) where \Gamma(z)=\int_{0}^{+\infty}t^{z-1}e^{-t}\mathrm{d}t , \alpha=2a/\sigma and \beta=2b/\sigma . The CIR model is very popular in many important applications such as interest rate modeling ( Cox, Ingersoll and Ross (1985) ; Lamberton and Lapeyre (1997) ), stochastic volatility stock markets Heston (1993) ; Berdjane and Pergamenshchikov (2013) ; Nguyen and Pergamenshchikov (2017) ) and, moreover, in Pergamenchtchikov, Tartakovsky and Spivak (2022) discrete versions of CIR processes are used in the epidemic analysis. To obtain reliable statistical inferences within this model, estimating the unknown parameters with guaranteed accuracy properties is necessary in the non-asymptotic setting. It should be noted that for the model of type ( 1 ) the usual maximum likelihood estimators are nonlinear functions of observations and it is not clear how to study such functions directly on the fixed time interval. For these reasons to overcome these dificulties in Ben Alaya, Ngô and Pergamenchtchikov (2025) ; Novikov and Shiryaev and Kordzakhia (2024) , sequential guaranteed estimation methods were developed to estimate the parameters of the model ( 1 ) with fixed estimation accuracy. Unfortunately, the proposed sequential procedures do not control the observation duration, which restricts their applications in many practical applications since, in practice, the duration of observation is usually bounded. For example, for the portfolio optimisation problems for financial markets with unknown parameters in Berdjane and Pergamenshchikov (2015) , it is shown that to construct optimal and robust financial strategies, it is necessary to use guaranteed truncated sequential estimators that can provide a fixed known estimation accuracy over fixed time intervals. For statistical models in discrete time truncated sequential procedures are developed in Konev and Pergamenshchikov (1990) and for the stochastic differential equations with the bounded diffusion coefficients such procedures were proposed in Konev and Pergamenshchikov (1992) ; Galtchouk and Pergamenshchikov (2011) ; Galtchouk and Pergamenshchikov (2015) ; Galtchouk and Pergamenshchikov (2022) for parametric and nonparametric problems. Unfortunately, these results can not be applied to stochastic differential equations with unbounded diffusion coefficients as, for example, ( 1 ). To study non-asymptotic estimation methods for such models, one needs to develop new analytic tools based on the special form of this process. The main goal of this paper is to develop truncated guaranteed estimation methods for the coefficients a and b , in the both scalar and two-dimensional parameter cases, on the basis of the observations (X_{\mathchoice{t}{t}{\lower 0.75346pt\hbox{$\scriptstyle t$}}{\lower 0.75346pt\hbox{$\scriptscriptstyle t$}}})_{\mathchoice{0\leq t\leq T}{0\leq t\leq T}{\lower 0.75346pt\hbox{$\scriptstyle 0\leq t\leq T$}}{\lower 0.75346pt\hbox{$\scriptscriptstyle 0\leq t\leq T$}}} of the process ( 1 ), where the observation duration T>0 is fixed in advance. 1.2 Main contributions In this paper, for the first time, the truncated sequential guaranteed methods were developed for the models ( 1 ) with unbounded diffusion coefficients. The proposed estimators for the parameters a and b have a guaranteed non-asymptotic mean square estimation accuracy, which is found in the explicit form. Moreover, through the asymptotic analysis methods developed in Ben Alaya, Ngô and Pergamenchtchikov (2025) it is shown that the proposed truncated guaranteed procedures are optimal in the minimax sense for local and general quadratic risks. For the local risks, the optimality properties are established in the class of all sequential and non sequential estimators with the observation duration less than T when T\to\infty . It is important to emphasize here that it is also established that the proposed truncated procedures asymptotically, without deteriorating the estimation quality, significantly reduce the observation period compared to usual non-sequential estimations. Moreover, for the general quadratic risk, the optimality properties are established in the class of all sequential procedures with mean observation time not exceeding the mean observation time of the proposed truncated procedures. It should be noted here that this class is sufficiently large since it includes all possible sequential procedures that can use more than T observations duration; only the mean observation duration has to be less than the mean observation duration of the proposed truncated procedures. This means that any sequential procedure having the same mean observation duration can not improve the accuracy properties with respect to the proposed one. 1.3 Organisation of the paper The rest of the paper is organized as follows. In Section 2 , we analyze the scalar truncated sequential estimation methods for the model ( 1 ). In Section 3 , we develop the two-step sequential estimation method for the parameter vector \theta=(a,b)^{\top} in the model ( 1 ). In Section 3.2 , we find conditions on the parameters of the process ( 1 ) which provide the optimality properties in minimax sense for the proposed sequential procedures. Section 4 presents the concentration inequalities for the CIR process. Some important conclusions are given in Section 5 . Appendix A contains some useful properties of the CIR process ( 1 ) and some auxiliary lemmas recalled from Ben Alaya, Ngô and Pergamenchtchikov (2025) . 2 Scalar truncated sequential procedures 2.1 Guaranteed estimation First, we consider the estimation problem for the parameter b in the process ( 1 ) in the case, when a

Read the complete text