Research

Questions, methods, artifacts, limitations

The common thread is data that arrives fragmented — duplicated, inconsistently described, split across sources and views, or unobserved below a measurement limit — and the question of how much structure can honestly be recovered from it.

Selected publications

Accepted / forthcoming

  1. Du, R., Chu, X., Jin, J., Luo, L., Lee, J. H., Hudson, L. G., MacKenzie, D. A., & Lewis, J. (accepted July 19, 2026). A data-driven single substitution to overcome bias from conventional single substitution for LOD missingness in covariates in least squares regression. Journal of Applied Statistics.

    Measurements below a limit of detection are often replaced by a fixed value before analysis. When the affected measurement is used as a covariate in least squares regression, that apparently simple preprocessing choice can distort the estimated relationship. This accepted work develops a data-driven single-substitution approach intended to reduce the bias associated with conventional single substitution while retaining the operational simplicity of a single completed analysis dataset. The study treats substitution as part of the statistical model rather than as a neutral housekeeping step.

    Status: Accepted on July 19, 2026; forthcoming. A DOI, final citation details, and any permitted manuscript link will be added after the publisher record is available.

Published

  1. Chu, X., Luo, X. R., & Chen, Y. (2019). A systematic review on cross-cultural information systems research. Information & Management, 56(3), 403–417.

Selected earlier work

  1. 张平康, 储新民, & 张仙锋 (2004). 电力市场化改革规律探索——从国际经验看我国的电力市场化改革. 山西财经大学学报 (Journal of Shanxi Finance and Economics University), 26(2), 73-76.

    An Exploratory Discussion on Rules Guiding Reform of the Power Market — Examining China's Power Market Reform in Light of International Experiences

  2. 储新民, & 李琪 (2009). 发展我国农业电子商务:制度重于技术——基于国内研究文献的述评与思考. 情报杂志 (Journal of Intelligence), 28(8), 6-9.

    Development of China's Agricultural e-commerce: Institution Priority to Technology — Thinking and Review Based on the Domestic Literature

  3. 储新民, 袁雪峰, 司增绰, 陈学法, & 李琪 (2011). 关于农业基础地位的实证研究——基于江苏与山东的比较. 经济问题 (On Economic Problems)(11).

    An empirical study of agriculture's foundational position: comparing Jiangsu and ShandongDraft translation

  4. 储新民, & 李琪 (2018). 乡村振兴背景下农民增收保障机制与政策研究——基于文献研究的分析. 淮阴工学院学报 (Journal of Huaiyin Institute of Technology), 27(4), 69-74.

    Research on increasing farmers' income under the Rural Revitalization Strategy: Analysis Based on Literature Research

All publications

Research areas

Each area states what is being asked, how it is approached, what exists publicly, and where the current limits are.

Computational Science — the interdisciplinary frame

What does it take to move between economics, statistics, and computation without losing the standards of any of them?

This is the frame the areas below sit inside rather than an additional research method. Doctoral work in computational science at the University of Massachusetts Boston, building on graduate training in statistics and economics, supplies the standards each area is held to: measurement taken seriously, inference under noise, and computation that survives contact with real data.

Methods

  • Statistical modeling and numerical methods
  • Machine learning applied to data-integration problems
  • Systematic literature review methodology

Public artifacts

Limitations. Interdisciplinary breadth is a bet: it trades depth-per-field for connections between fields, and the copy here claims only the work actually done.

Status: Interdisciplinary frame; ongoing

Entity Resolution and Deduplication

When do two records describe the same real-world thing — and how do we make that decision auditable as sources multiply and volume grows?

Matching and deduplicating records across noisy, heterogeneous sources, with pipelines whose every stage can be inspected, replaced, compared, and audited. The pipeline is treated as the object of study rather than as plumbing around a matcher: ingestion, blocking, comparison, and clustering are separately measurable, so a change in one stage can be attributed rather than merely observed. The questions that follow are where such pipelines degrade as volume and source heterogeneity grow, and what has to be recorded for a merge decision to remain reconstructible afterwards.

Methods

  • Multi-field similarity combining string, numeric, and embedding-based metrics
  • End-to-end pipelines from ingestion to evaluation (erbot toolkit)
  • Replaceable, separately measurable stages so effects can be attributed
  • Reproducible evaluation binding every metric to the configuration that produced it
  • Automated parameter tuning with recorded configurations

Public artifacts

Limitations. No benchmark figures are published in isolation; results are only meaningful with their dataset and evaluation setting.

Status: Active

Record Linkage and Blocking

How can candidate pairs be generated so that linkage stays tractable without silently discarding true matches?

Blocking and candidate-generation strategies that control the quadratic comparison budget while keeping recall losses measurable rather than invisible.

Methods

  • Blocking utilities integrated into the erbot pipeline
  • Recall-aware analysis of candidate generation choices

Public artifacts

Limitations. Blocking trade-offs are dataset-dependent; conclusions are reported per setting, not as universal rules.

Status: Active

Multiview Representation and Fusion

How should multiple views of the same records — lexical, numeric, embedding, tensor — be combined without letting one view dominate or contradict the rest?

Representing records through several complementary views and fusing them, including tensor-based canonical correlation approaches for aligning views.

Methods

  • TF–IDF with singular value decomposition alongside embedding features
  • Tensor-based multiple canonical correlation analysis (tensorMCCA)
  • Contributions to R packages for multiview analysis (GCMER)

Public artifacts

Limitations. View-fusion behavior under highly imbalanced view quality remains an open question in this work.

Status: Active

Similarity Graphs and Graph Clustering

Once pairwise similarities exist, how should entities be formed — and which clustering assumptions quietly change the answer?

Turning similarity structure into entity clusters, comparing K-means, Louvain community detection, hierarchical clustering, and partitioning around medoids on the same footing.

Methods

  • Similarity-graph construction from multi-field comparisons
  • Side-by-side clustering strategies within one pipeline (K-means, Louvain, hierarchical, PAM)

Public artifacts

Limitations. Cluster-quality comparisons depend on ground-truth availability, which many real datasets lack.

Status: Active

Threshold Calibration

Where should a match threshold sit when the cost of a false merge and a missed match are not symmetric — and how stable is that choice?

Systematic threshold selection for match decisions: optimization curves, sensitivity to data shifts, and recording the configuration behind every operating point.

Methods

  • Automated parameter tuning with optimization curves in evaluation reports
  • Sensitivity analysis of operating points across settings

Public artifacts

Limitations. Calibrated thresholds transfer poorly across domains; per-domain recalibration is assumed, not avoided.

Status: Active

Evaluation Integrity and Reproducibility

Can an entity-resolution result be trusted — and rerun — by someone who was not in the room when it was produced?

Making evaluation a first-class artifact: ground-truth comparison, reports that bind metrics to the exact parameters that produced them, and pipelines designed to be rerun.

Methods

  • Evaluation against ground truth built into the pipeline
  • Generated PDF reports carrying parameters, optimization curves, and metrics together

Public artifacts

Limitations. Reproducibility guarantees stop at the data boundary: private or shifting source data limits what any pipeline can promise.

Status: Active

Missing Data and Regression

How do substitutions for values below a limit of detection alter least squares regression, and how can a simple completed-data workflow be made less biased?

This work examines missingness created by measurement limits, with particular attention to the inferential consequences of substituting values in covariates. The emphasis is not only on filling an unavailable value, but on understanding how that choice propagates into regression estimates.

Methods

  • Least squares regression under limit-of-detection missingness
  • Bias analysis for covariate substitution
  • Data-driven construction and evaluation of a single-substitution rule

Public artifacts

  • Du, Chu, Jin, Luo, Lee, Hudson, MacKenzie & Lewis — accepted by the Journal of Applied Statistics (July 19, 2026); see Selected publications above

Limitations. The public website currently reports the accepted status and research question only. Detailed methodological and numerical claims must remain tied to the accepted manuscript and eventual publisher record.

Status: Accepted paper; ongoing related work

Research directions in development

Directions being developed alongside the established areas. Each states the question it asks and what it would examine. None reports results, because none has produced any yet.

Tensor and multiview data integration

How can several heterogeneous measurement modalities of the same subjects be combined so that shared structure is recoverable without any one modality dominating the joint representation?

A continuation of the established Multiview Representation and Fusion work rather than a separate programme: it carries tensor-based multiple canonical correlation analysis into settings where the modalities describing the same subjects differ in kind, in dimensionality, and in measurement quality.

Without appropriate scaling, weighting, or regularization, differences in variance, dimensionality, and measurement quality can cause one modality to dominate a joint representation. The direction is concerned with the estimation procedure and with the conditions under which recovered factors stay interpretable, rather than with any single application.

Under consideration

  • Tensor decomposition and tensor multiple canonical correlation analysis
  • Identifiability and estimation stability under unequal modality quality
  • Criteria for deciding when a recovered factor is interpretable

Relation to established work. An extension and continuation of the established Multiview Representation and Fusion work, not a separate programme.

Status: Research Direction in Development

Constraint-aware graph formulations for entity resolution

How should conflicting pairwise match decisions be resolved coherently when a graph formulation makes the conflicts explicit rather than hiding them behind a threshold?

Pairwise match scores routinely disagree with one another. This direction examines graph formulations that represent positive linkage evidence and incompatibility constraints coherently, so that a conflict between two match decisions becomes an explicit object to be resolved rather than an artefact suppressed by a cut-off.

The work is at the stage of settling what the graph should represent. Which construction follows from that choice is the question, not the premise.

Under consideration

  • Graph construction from multi-field comparisons under explicitly stated edge semantics
  • Representing positive linkage evidence and incompatibility constraints together
  • Evaluation against conventional threshold-based baselines

Relation to established work. A methodological extension of Similarity Graphs and Graph Clustering.

Status: Research Direction in Development

Cost-sensitive and auditable entity resolution

How should record linkage be evaluated when a false link and a missed link carry very different costs, and the result has to be defensible to a human reviewer?

An applied extension of the established entity-resolution work. In many settings the two error types are not interchangeable, and an evaluation that reports a single accuracy figure hides the trade-off that actually matters.

The direction treats asymmetric error cost, auditability, and privacy exposure as first-order evaluation criteria rather than as considerations added afterwards, and asks how accuracy can be estimated at all when ground truth is partial.

Under consideration

  • Asymmetric false-link and missed-link costs
  • Estimation under partial ground truth
  • Privacy-conscious evaluation protocols
  • Auditability of individual merge decisions
  • Human review as part of the design, not an afterthought
  • Constraints that regulated settings place on evaluation

Relation to established work. An applied extension of Entity Resolution and Deduplication.

Status: Research Direction in Development