CoChat
Sign Up
Cited by 898Statistical Methods in EpidemiologyAdvanced Causal Inference TechniquesStatistical Methods and Inference

Sparse data bias: a problem hiding in plain sight

Open access
Greenland, Sander, Mansournia, Mohammad Ali, Altman, Douglas G·2016·BMJ
DOI 10.1136/bmj.i1981
Open this paper in CoChat — free accountOpen PDF

Abstract

Effects of treatment or other exposure on outcome events are commonly measured by ratios of risks, rates, or odds. Adjusted versions of these measures are usually estimated by maximum likelihood regression (eg, logistic, Poisson, or Cox modelling). But resulting estimates of effect measures can have serious bias when the data lack adequate case numbers for some combination of exposure and outcome levels. This bias can occur even in quite large datasets and is hence often termed sparse data bias. The bias can arise or be worsened by regression adjustment for potentially confounding variables; in the extreme, the resulting estimates could be impossibly huge or even infinite values that are meaningless artefacts of data sparsity. Such estimate inflation might be obvious in light of background information, but is rarely noted let alone accounted for in research reports. We outline simple methods for detecting and dealing with the problem focusing especially on penalised estimation, which can be easily performed with common software packages.

The claims and statistics of this paper haven’t been checked yet.

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

A few free checks a day · papers already verified open instantly · results are public

Integrity signals

Registry

Crossref · 10.1136/bmj.i1981

Cites

39 of 39 references verified0 citation errors · 0 not checkable

Statistics

Claims the paper makes

The claims and statistics of this paper haven’t been checked yet.

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

A few free checks a day · papers already verified open instantly · results are public

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

  • There is a sparse data risk in research with transgender individuals because of small sample sizes and short study durations.

    “there is a sparse data risk66 because of small sample sizes and short study durations”

    From How does hormone transition in transgender women change body composition, muscle strength and haemoglobin? Systematic review with a focus on the implications for sport participation (2021) · cites this paper as [66]

    The Abstract (page unavailable) explains that sparse data bias can arise when there are too few cases in combinations of exposure and outcome categories, supporting the concern about small samples. The paper does not address research with transgender individuals or identify short study durations as a cause of sparse data bias. (LLM, conf=0.86)

    Experimental — accuracy not yet measured

  • The pooled OR estimates of this analysis may be limited by sparse data bias, which can inflate estimates of effect measures and should be considered when interpreting the findings.

    “This analysis may also be limited by the possibility of sparse data bias in the pooled OR estimates, as indicated by the unconventional ORs and wide confidence limits produced by some studies.This is problematic as such bias can inflate the estimate of effect measures, which should be taken into account when interpreting the findings of the present analysis.37”

    From Risk of atrial fibrillation in athletes: a systematic review and meta-analysis (2021) · cites this paper as [37]

  • Confounder imprecision does not bias the target estimator, that is, the association of physical activity on severe COVID-19 outcomes.

    “However, confounder imprecision does not bias the target estimator, that is, the association of PA on severe COVID-19 outcomes. 29”

    From Physical inactivity is associated with a higher risk for severe COVID-19 outcomes: a study in 48 440 adult patients (2021) · cites this paper as [29]

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

The self-check needs the paper’s full text.

Cites

Flagged first, then by citations

Research works printed but not linked: 1

  1. Glymour MM, Greenland S. Causal diagrams. In: Rothman KJ, Greenland S, Lash TL, eds. Modern Epidemiology.3rd ed. Lippincott Williams & Wilkins, 2008: 183-209.

Cited by

Sorted by citations

Related

Authors
Greenland, S., Mansournia, M. A., & Altman, D. G.
Affiliations
  • University of California, Los Angeles (US)
  • University of Oxford (GB)
ORCID
Venue
BMJ
Publisher
BMJ
Published
2016-04-27
Volume
352
Pages
i1981
ISSN
1756-1833 (ISSN-L)
DOI
10.1136/bmj.i1981
OpenAlex
W2343557095
Semantic Scholar
30977540a580178337ada4e9b42e3df884bb51ce
Cited by
898
References
38
Citations by year
2026: 77 · 2025: 83 · 2024: 92 · 2023: 94 · 2022: 82 · 2021: 98 · 2020: 99 · 2019: 80 · 2018: 75 · 2017: 109 · 2016: 8
FWCI
91.86
Type
journal article
Language
en
Open access
Licence
www.bmj.org/licenses/tdm/1.0/terms-and-conditions.html (tdm)
Locations
  • BMJ (publishedVersion)
  • PubMed (publishedVersion)
  • HighWire Press Open Archive (submittedVersion)
Topics
  • Physical Sciences › Mathematics › Statistics and Probability › Statistical Methods in Epidemiology
  • Physical Sciences › Mathematics › Statistics and Probability › Advanced Causal Inference Techniques
  • Physical Sciences › Mathematics › Statistics and Probability › Statistical Methods and Inference
Keywords
relative risk, odds ratio, rate ratios, confounding adjustment
MeSH terms
Humans, Bias, Likelihood Functions, Odds Ratio, Outcome Assessment, Health Care, Sample Size, Risk Assessment, Quality Improvement

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review