Browsing by Author "Ghachem, M."

Now showing 1 - 2 of 2

Citation - WoS: 2
Citation - Scopus: 2
Estimation of the Probability of Informed Trading Models Via an Expectation-Conditional Maximization Algorithm
(Springer Science and Business Media Deutschland GmbH, 2025) Ghachem, M.; Ersan, O.
The estimation of the probability of informed trading (PIN) model and its extensions poses significant challenges owing to various computational problems. To address these issues, we propose a novel estimation method called the expectation-conditional-maximization (ECM) algorithm, which can serve as an alternative to the existing methods for estimating PIN models. Our method provides optimal estimates for the original PIN model as well as two of its extensions: the multilayer PIN model and the adjusted PIN model, along with its restricted versions. Our results indicate that estimations using the ECM algorithm are generally faster, more accurate, and more memory-efficient than the standard methods used in the literature, making it a robust alternative. More importantly, the ECM algorithm is not limited to the models discussed and can be easily adapted to estimate future extensions of the PIN model. © The Author(s) 2025.
A Methodological Approach to the Computational Problems in the Estimation of Adjusted PIN Model
(Routledge, 2025) Ersan, O.; Ghachem, M.
It is well documented that computational problems may lead to large biases in the estimation of probability of informed trading (PIN) models. The complexity of the AdjPIN model [Duarte, J. and Young, L., Why is PIN priced? J. Financ. Econ., 2009, 91, 119–138.], an extension of the conventional PIN model, exacerbates further these computational issues due to its larger parameter set. We introduce a dual approach to improve estimation reliability: a logarithmic factorization of the likelihood function and a strategic algorithm for generating initial parameter sets. The logarithmic factorization addresses floating point exceptions and numerical instability, while the algorithm significantly reduces the likelihood of converging to local maxima. We show that our methodology outperforms existing best practices and it enables accurate estimation of the AdjPIN model. We, therefore, strongly suggest its use in future studies. © 2025 Informa UK Limited, trading as Taylor & Francis Group.