Nonparametric estimation and testing in semiparametric autoregressive conditional duration models

Pipat Wongsaart

    Research output: ThesisDoctoral Thesis

    175 Downloads (Pure)


    The advent of the so-called transaction data in finance has given econometrician the tool to address a variety of issues surrounding the structure of the trading process and/or price discovery in nancial markets. However, transaction data pose a number of unique econometric challenges that do not easily fit into the traditional modeling framework that have been developed so far in the literature. The ultimate goal of this thesis is to establish a novel econometric method of estimating the conditional intensity of the arrival times of financial events. This goal can be broken down into a few research objectives. (1) Firstly, it is to establish a new generation (semiparametric) approach to efficiently model the dynamics of the waiting time between the arrivals of financial events or what is commonly known as duration. (2) Secondly, it is to derive a set of estimators, so that empirical estimates of the density, survival and the baseline intensity functions associated with duration processes can be calculated. (3) Thirdly, it is to develop a novel testing procedure to test the marginal density function of financial durations. While the first and second objectives are discussed in detail in Chapter 2, the third objective is considered in Chapter 3. These semiparametric estimation and nonparametric testing procedure are introduced in conjunction with the detailed theoretical and experimental examinations of their statistical validity. Furthermore, the usefulness and practicability of these methods are illustrated using various datasests from both foreign exchange and international stock markets.
    Original languageEnglish
    QualificationDoctor of Philosophy
    Publication statusUnpublished - 2011


    Dive into the research topics of 'Nonparametric estimation and testing in semiparametric autoregressive conditional duration models'. Together they form a unique fingerprint.

    Cite this