HZHaibin Zhu

Haibin Zhu

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


Haibin Zhu has a total of 14 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Intraday Networks: Estimation and Community Inference 

    by Haibin Zhu, Wenhao Cui, Zhenggang Wang

    We develop methods for estimating intraday networks from binary event observations and testing changes in their community structure over a fixed time span. We first establish max-norm consistency of local estimators and a stable central limit theorem for their fixeddimensional contrasts. Under a degree-corrected block model, row-normalized, unscaled signal eigenvectors yield a co-membership representation whose population value is invariant to activity and connectivity. We prove uniform community recovery on pure blocks and consistency of integrated co-membership, allowing supplied or consistently estimated counts to vary across regimes. For inference, we propose an event-level multiplier bootstrap, which provides asymptotically valid tests and simultaneous confidence bands for co-membership contrasts while accounting for dependence within event vectors and across shared blocks. Simulations illustrate the finite-sample performance of the proposed methods. Applications to stock price staleness compare local network weights and fitted groups around an exchange disruption and scheduled monetary policy announcements.

  • Open Access

    A Nonparametric Test for Intraday Variation in Systematic Price Staleness

    by Yuexi Zhao, Wenjing Liu, Haibin Zhu

    We propose a nonparametric test for intraday variation in systematic price staleness. We view systematic price staleness as a market-wide component in very small price adjustments across assets. The main statistical test assesses whether systematic staleness remains constant within a trading day after removing a regular time-of-day pattern. Under infill and long-span asymptotics, the statistic is asymptotically standard normal under the null and consistent against fixed alternatives. Simulations show good size and power. Using oneminute data for nine U.S. sector SPDR ETFs from 2020 to 2025, the raw test rejects on 99.92% of regular trading days, while the diurnally corrected test rejects on 12.72%; 9.50% remain significant after a false-discovery rate control. Corrected rejections are more common during market stress and policy-event windows and on days with higher cross-ETF correlation, large market moves, and higher trading volume. The test provides a simple diagnostic for intraday patterns in market-wide price staleness.

  • Open Access

    Forecasting Realized Volatility Intervals with Pooled Conformal Calibration

    by Jing Li, Yiming Liu, Ziyu Wang, Haibin Zhu

    This paper develops a pooled conformal calibration method for realized volatility panels. Standard conformal methods construct prediction intervals asset by asset and therefore do not use cross-sectional information in large financial panels. This can lead to simultaneous under-coverage when many assets are exposed to the same market-wide shock. We propose Pooled Conformal Prediction for Realized Volatility (PCP-RV), a simple calibration method that combines time-weighted conformal prediction with cross-sectional pooling of standardized forecast errors. For each asset, PCP-RV uses the larger of the asset-specific time-weighted margin and the pooled systematic margin to construct the prediction interval. Under a volatility factor structure, the pooled margin can be viewed as a buffer against common forecast uncertainty. Simulations show that our method improves coverage in settings with persistent common shocks and regime changes. In empirical applications, PCP-RV delivers coverage closer to the nominal level and improves cross-sectional coverage reliability across several point forecasting models. We also show how the resulting intervals can be used for intraday risk monitoring and Value-at-Risk measurement.

  • Open Access

    Decomposing High-Frequency Order Flow: Commonality and Idiosyncratic Trade Imbalance

    by Wenhao Cui, Haibin Zhu

    Standard measures of trade imbalance often fail to distinguish between asset-specific directional pressure and market-wide trading activity, leading to misleading inferences. This paper develops a continuous-time framework to decompose high-frequency order flow into common and idiosyncratic components using bivariate Cox processes with a latent common factor. We propose a novel Idiosyncratic Trade Imbalance (IdTI) index, designed to filter out interference from common factors. To facilitate inference, we establish a coupling result using a fixed number of observations that supports Poisson approximations for high-frequency data and develop a Bayesian procedure with asymptotic validity. Simulation results indicate that traditional frequentist methods become unreliable during market-wide intensity shocks, while our proposed methodology maintains robust coverage. Applying the framework to tick-by-tick data during major product launches and FOMC announcements, we find that the idiosyncratic measure unmasks significant directional convictions that are otherwise obscured by common trade intensity.

  • Open Access

    Fixed-k Inference for Spot R-squared in High-frequency Factor Models 

    by Haibin Zhu

    We introduce a novel fixed-k econometric framework for nonparametric inference on the instantaneous coefficient of determination, the spot R-squared, in continuous-time factor models. Unlike conventional large-k asymptotics, which suffer from smoothing bias and rely on unstable nuisance parameters, we formulate local regression as a finite-sample Gaussian limit experiment. We show that the spot R-squared estimator converges to a random variable with a hierarchical noncentral F distribution. This pivotal limit theory facilitates exact local inference that is completely free of nuisance parameters and asymptotically Uniformly Most Powerful Invariant (UMPI). Furthermore, we establish uniform confidence bands for the entire spot R-squared trajectory by coupling the non-Donsker functional limit with extreme value theory. Illustrating the practical utility of our methodology, we formalize a high-frequency Regression Discontinuity Design (RDD) to test for structural breaks around FOMC announcements and document abrupt, market-wide algorithmic herding triggered by macroeconomic shocks.

  • Open Access

    Library size-stabilized metacells construction enhances co-expression network analysis in single-cell data

    by Tianjiao Zhang, Haibin Zhu

    Single-cell RNA sequencing (scRNA-seq) deciphers cell type-specific co-expression networks to resolve biological functions but remains constrained by data sparsity and compositional biases. Conventional metacells construction strategies mitigate sparsity by aggregating transcriptionally similar cells but often neglect systematic biases introduced by compositional data. This problem leads to spurious co-expression correlations and obscuring biologically meaningful interactions. Through mathematical modeling and simulations, we demonstrate that uncontrolled library size variance in traditional metacells inflates false-positive correlations and distorts co-expression networks. Here, we present LSMetacell (Library Size-stabilized Metacells), a computational framework that explicitly stabilizes library sizes across metacells to reduce compositional noise while preserving cellular heterogeneity. LSMetacell addresses this by stabilizing library sizes during metacells aggregation, thereby enhancing the accuracy of downstream analyses such as Weighted Gene Co-expression Network Analysis (WGCNA). Applied to a postmortem Alzheimer’s disease brain scRNA-seq dataset, LSMetacell revealed robust, cell type-specific co-expression modules enriched for disease-relevant pathways, outperforming the conventional metacells approach. Our work establishes a principled strategy for resolving compositional biases in scRNA-seq data, advancing the reliability of co-expression network inference in studying complex biological systems. This framework provides a generalizable solution for improving transcriptional analyses in single-cell studies.

  • On Bivariate Time-Varying Price Staleness

    by Haibin Zhu, Zhi Liu

    Price staleness refers to the extent of zero returns in price dynamics. Bandi et al. (2020c) introduce two types of staleness: systematic and idiosyncratic staleness. In this study, we allow price staleness to be time-varying and study the statistical inference for idiosyncratic and common price staleness between two assets. We propose consistent estimators for both time-varying idiosyncratic and systematic price staleness and derive their asymptotic theory. Moreover, we develop a feasible nonparametric test for the simultaneous constancy of idiosyncratic and common price staleness. Our inference is based on infill asymptotics. Finally, we conduct simulation studies under various scenarios to assess the finite sample performance of the proposed approaches and provide an empirical application of the proposed theory.


  • Open Access

    A two‐stage neural network prediction of chronic kidney disease

    by Hongquan Peng, Haibin Zhu, Chi Wa Ao Ieong, Tao Tao, Tsung Yang Tsai, Zhi Liu

    Accurate detection of chronic kidney disease (CKD) plays a pivotal role in early diagnosis and treatment. Measured glomerular filtration rate (mGFR) is considered the benchmark indicator in measuring the kidney function. However, due to the high resource cost of measuring mGFR, it is usually approximated by the estimated glomerular filtration rate, underscoring an urgent need for more precise and stable approaches. With the introduction of novel machine learning methodologies, prediction performance is shown to be significantly improved across all available data, but the performance is still limited because of the lack of models in dealing with ultra‐high dimensional datasets. This study aims to provide a two‐stage neural network approach for prediction of GFR and to suggest some other useful biomarkers obtained from the blood metabolites in measuring GFR. It is a composite of feature shrinkage and neural network when the number of features is much larger than the number of training samples. The results show that the proposed method outperforms the existing ones, such as convolutionneural network and direct deep neural network.