Haseman–Elston regression

A statistical-genetics regression of a declared pairwise quantitative-trait response on a matched pairwise genetic-sharing measure, with slope meaning conditional on the study design and genetic model.

Version

v1 · 2026-10-31 · History

Domain-specific #

14508

Domain group

Natural Sciences

Origin domain

Biology & Ecology

Subdomains

Pairwise Regression, Linkage Analysis → Biology & Ecology

Core Idea

Haseman–Elston regression relates a quantitative-trait response constructed for a pair of people to a genetic-sharing measure for the same pair. The original construction squares full siblings' trait difference and regresses it on estimated marker identity by descent (IBD). A slope is evidence about linkage only under the stated genetic and marker model; its sign alone does not identify a causal locus or recombination fraction.^[1]

How would you explain it like I'm…

Sibling Gene-Sharing Check

Brothers and sisters share some pieces of their DNA, but not always the same amount. Scientists check: when two siblings share more of one DNA piece, are they more alike in something like blood pressure? If so, that's a clue the piece is near something that matters — a clue, not proof.

Siblings, Shared Genes, Similar Traits

Haseman–Elston regression is a method that looks at pairs of people. In the original version, from 1972, scientists took pairs of full brothers and sisters. For each pair, they found how different the two siblings were in a measured trait and squared that difference. Then they checked whether pairs who shared more DNA at a marker spot, passed down from the same parents, had smaller differences. A relationship like that can be a clue that a gene affecting the trait is nearby. Later versions changed things, like using unrelated people, but not every study on traits counts as this method.

Pairwise Trait–Sharing Regression

Haseman–Elston regression relates a trait measure computed for a pair of people to a measure of genetic sharing for the same pair. The original 1972 method used full siblings: it squared the difference in their values of a quantitative trait and regressed that on the estimated proportion of marker alleles they share identical by descent (IBD). Under the original genetic and marker model, the slope can be evidence of linkage between the marker and a gene affecting the trait, but its sign alone doesn't identify the causal gene or the recombination fraction. The name also covers later adaptations that change the pairs or responses — for instance covariate-adjusted sibling analyses of blood pressure, or versions using unrelated people and identity-by-state sharing from SNPs. Not every trait regression or heritability estimate is Haseman–Elston regression.

 

Haseman–Elston regression regresses a quantitative-trait response constructed for a pair of individuals on a genetic-sharing measure for that pair. The original 1972 formulation used full-sibling pairs, taking the squared trait difference as the response and the estimated proportion of marker alleles shared identical by descent (IBD) as the predictor. Under the original genetic and marker model, the fitted slope provides evidence about linkage between the marker and a trait locus. Its sign alone does not identify a causal locus or estimate the recombination fraction. The name extends to explicitly described later adaptations that declare changed pair and response conventions. Wu and colleagues compared standard and covariate-adjusted sibling-pair analyses of observed Framingham systolic blood pressure. Chen described an adaptation for unrelated individuals using SNP-based identity-by-state (IBS) relatedness, illustrated with a simulation. These are applications of a pairwise trait–sharing regression; not every regression on a phenotype, and not every heritability estimator, is Haseman–Elston regression.

Structural Signature

• Pair design: identify the pairs and the population or sampling convention.
• Matched response: calculate one declared quantitative-trait response for each pair.
• Sharing predictor: calculate a genetic-sharing measure for that same pair.
• Qualified regression: fit the response on the predictor and interpret its slope under the declared study and genetic model.^[1][2][3]

What It Is Not

A generic regression without a pairwise genetic-sharing predictor does not instantiate this method. Later adaptations can use different pairs, responses or sharing measures; they do not make all variants interchangeable. Pairs can share individuals, and uncertainty depends on the analysis rather than an automatic independence assumption.^[1][2][3]

Examples

Observed sibling data. Wu and colleagues analyzed Framingham systolic blood pressure, comparing a standard response based on visit averages with an adjusted response built from common visits. These are two response constructions within one observed study, not separate cohorts.^[2]

Unrelated-person simulation. Chen simulated pairs of unrelated individuals and used SNP identity-by-state-derived relatedness. This illustrates a variant under the paper's simulation model, not a general empirical calibration guarantee.^[3]

Reasoning and Boundary

State how pairs, the trait response and the sharing predictor are constructed before fitting the regression. Distinguish marker IBD from SNP IBS-derived relatedness, and interpret the fitted slope using the relevant model and uncertainty analysis. A linkage result is not itself identification of a causal gene.^[1][2][3]

Neighborhood in Abstraction Space

Haseman–Elston regression sits in a sparse region of the domain-specific corpus (94^th percentile for distinctiveness): few abstractions share its structure, so a faithful description tends to retrieve it precisely.

Family — Population Genetics & Selection (13 abstractions)

Nearest neighbors

• Disassortative mating — 0.79
• Generalist Genes Hypothesis — 0.78
• Quantitative Trait Locus Mapping — 0.77
• Yoked Control Design — 0.77
• Co-adaptation — 0.77

Computed from structural-signature embeddings · 2026-10-31

References

[1] Haseman, J. K., and Elston, R. C. (1972). The investigation of linkage between a quantitative trait and a marker locus. Behavior Genetics 2(1):3–19. Full original PDF: one-based PDF p. 3 / printed p. 5 §1 for the squared-pair response; PDF p. 5 / printed p. 7 §1 Eqs. (8)–(9) and §2 Eq. (11) for regression and estimated marker IBD; PDF p. 7 / printed p. 9 §3 Eqs. (13)–(14) and PDF pp. 10–11 / printed pp. 12–13 for model-dependent slope and limits. The Introduction's Eq. (1) is an individual-trait model, not the pairwise regression. The available PDF's OCR corrupts Greek symbols and superscripts; no exact original-symbol equation is reproduced here. registry ↩a ↩b ↩c ↩d

[2] Colin O. Wu, Gang Zheng, Eric Leifer, Dean Follmann, and Jing-Ping Lin (2003). A statistical method for adjusting covariates in linkage analysis with sib pairs. BMC Genetics 4(Suppl 1):S51. Full original publisher article, Methods for longitudinal studies (retrieval lines 116–130), Implementation (lines 157–195), Results (lines 199–218) and Fig. 1. The article analyzes observed Framingham GAW13 sibling data and distinguishes the standard response from the common-visit adjusted response. registry ↩a ↩b ↩c ↩d

[3] Guo-Bo Chen (2014). Estimating heritability of complex traits from genome-wide association studies using IBS-based Haseman–Elston regression. Frontiers in Genetics 5:107. Full original publisher article, Theory (retrieval lines 413–421, 445–448), Monte Carlo Simulation I (lines 579–602 and Table 7), and Discussion (lines 730–735); publisher PDF. The featured unrelated-person case is simulated. Nearby Simulation I prose has slope-sign and number inconsistencies relative to Table 7, so no such numerical slope is quoted here. registry ↩a ↩b ↩c ↩d