Antrometric v2.1 for Precision Stature Estimation

A forensic tool for estimating human stature from skeletal long bones, developed at the International Institute of Forensic Expertise and available at antrometric.com

The question arrives with every set of fragmentary skeletal remains, whether on an excavation in Southeast Asia, in a laboratory processing disaster victim identification cases, or at a crime scene where the body has been found too late for soft tissue to tell the story. How tall was this person? The answer is not decorative. It is a biometric anchor that ties a set of bones to a specific human being, one of the few measurements that can be extracted from skeletal material and directly cross-referenced against the records of the missing, the disappeared, and the unidentified. Getting it right matters in a very concrete sense, and the traditional tools available to forensic practitioners, pocket calculators with published regression tables, manual interpolation across multiple formulas, and the inevitable transcription errors that come with field conditions, have not kept pace with what the science of stature estimation can actually deliver.

Antrometric v2.1 is the answer I built to that gap, and in this article I want to explain what it does, why the mathematical approach matters, and what the next version is already working toward.

Why Stature Estimation Needs a Dedicated Tool

Forensic anthropologists, archaeologists, criminalists, pathologists, and forensic medical examiners work across environments that share one characteristic: the need for a reliable stature estimate from incomplete skeletal material, often quickly, often without stable laboratory infrastructure, and occasionally on a smartphone screen in poor light. The regression formulas that have formed the scientific backbone of stature estimation since the foundational work of Mildred Trotter and Goldine Gleser in the 1950s are mathematically correct and rigorously validated. The challenge has never been the equations themselves, it has been the infrastructure for applying them consistently, accurately, and across the population diversity of real casework, which spans European, African, Asian, and Thai populations in the IIFE’s operational reach and extends further in international collaboration cases.

The approach Trotter and Gleser established, and which Antrometric implements in its current form, rests on linear regression models that predict stature from individual long bone measurements. For each bone, stature is computed from the equation

[math]\text{stature} = a \times \text{bone length} + b[/math]

where the slope a and intercept b are specific to the combination of bone type, sex, and ancestry, covering all six long bones of the appendicular skeleton: femur, tibia, fibula, humerus, radius, and ulna. This makes the tool directly applicable to the full range of practitioners who work with skeletal material, whether archaeologists documenting the demographic profile of a Bronze Age burial population, physical anthropologists reconstructing the stature distribution of a historical cohort, forensic practitioners identifying unknown individuals from fragmentary crime scene remains, or investigative authorities processing mass disaster victim identification cases where speed and accuracy are simultaneously critical (Trotter and Gleser, 1952, Estimation of stature from long bones of American Whites and Negroes, American Journal of Physical Anthropology, 10(4), 463-514; Trotter and Gleser, 1958, A re-evaluation of estimation of stature based on measurements of stature taken during life and of long bones after death, American Journal of Physical Anthropology, 16(1), 79-123). These coefficients have been validated across decades of forensic anthropological work, extended to additional population groups by subsequent researchers, and incorporated into the Feldesman et al. (1990, The femur/stature ratio and estimates of stature in mid and late-Pleistocene fossil hominids, Journal of Forensic Sciences, 35(2), 431-446) dataset that underpins the application’s European, African, and Asian modules. Thai population data derives from Mahakkanukrauh et al. (2011, Stature estimation from long bone lengths in a Thai population, Forensic Science International, 210(1-3), 279.e1-279.e7), based on 200 skeletons from Northern Thailand, whose inclusion reflects the practical reality that a significant share of IIFE’s international casework originates in Southeast Asian jurisdictions.

What a static regression table cannot do, and what Antrometric does, is combine measurements from multiple bones into a single, properly weighted estimate that systematically reduces the margin of error with each additional measurement, and present that estimate with a formally calculated combined uncertainty rather than leaving the practitioner to mentally integrate values from three separate formulas.

The Mathematics of Uncertainty Reduction

The core of the multi-bone estimation approach is the weighted mean, a statistical technique that assigns each individual bone estimate an influence on the final result proportional to its precision. A bone whose regression model carries a low standard error of estimate contributes more to the combined result than one with higher uncertainty. The weight assigned to each bone’s estimate is the reciprocal of its squared uncertainty,

[math]\text{weight}_i = \frac{1}{u_i^2}[/math]

where u_i is the standard error of the regression for that bone type within the relevant demographic group. The weighted stature estimate is then computed as

[math]\text{weighted stature} = \frac{\sum (\text{stature}_i \times \text{weight}_i)}{\sum \text{weight}_i}[/math]

and the combined uncertainty of the resulting estimate is derived as

[math]\text{combined uncertainty} = \frac{1}{\sqrt{\sum \text{weight}_i}}[/math]

The consequence of this approach is mathematically straightforward: as the number of measured bones increases, the sum of weights increases, and the reciprocal of its square root decreases. The combined uncertainty shrinks. Each additional bone drives the estimate toward greater precision, even when the additional bones individually carry higher standard errors than the first.

A concrete example makes this visible. A Thai male skeleton yields a femur measurement of 45 centimeters. The femur regression for this population group produces a stature estimate of 169.93 centimeters with a standard error of 5.06 centimeters, which means the weight assigned to this estimate is approximately 0.039. Adding a tibia measurement of 35 centimeters, which yields a stature of 165.10 centimeters with a standard error of 5.28 centimeters and a corresponding weight of approximately 0.036, changes the result significantly. The weighted combined stature works out to approximately 167.6 centimeters, and the combined uncertainty falls to approximately 3.65 centimeters, a reduction of more than a full centimeter from the femur estimate alone. In forensic terms, this is the difference between excluding and not excluding a missing person from a possible identification. The reduction is not incidental, it is the designed output of the weighted mean approach, and it continues to compound as additional bones enter the calculation.

The application performs these calculations in real time across all measured bones, presenting the combined estimate and its uncertainty immediately as each measurement is entered. Both centimeters and inches are supported with automatic conversion, reflecting the operational reality that international casework involves practitioners trained in different measurement systems who need their results in the units they will use for reporting.

Handling Unknown Sex Without Abandoning Precision

Sex determination from skeletal remains is possible when morphological indicators are preserved, but it is not always possible, and it is never simple. Juvenile remains, fragmentary material from mass disaster scenarios, highly degraded archaeological specimens, and cases where preservation has obscured the morphological markers that would otherwise guide the assessment all present the same problem: a regression formula requires a sex assignment, and the available material does not provide one.

The approach Antrometric takes is to average the male and female regression coefficients for the relevant ethnic group, producing a combined estimate that represents the midpoint of the sex-specific stature distributions for that population. The rationale follows directly from Trotter and Gleser’s (1958) foundational work, which established the sex-specific regression models but also documented the distribution of male and female stature within each population group. When sex is unknown, the averaged coefficients produce an estimate whose central value falls approximately where a randomly selected individual from the combined male and female distribution would be expected to fall.

To illustrate: a Thai individual of unknown sex with a femur length of 45 centimeters would yield, under the male regression, a stature of 169.93 centimeters with an uncertainty of 5.06 centimeters, and under the female regression, a stature of 166.20 centimeters with an uncertainty of 5.21 centimeters. Averaging the coefficients directly produces combined values of a = 2.46 and b = 57.37, which applied to the 45-centimeter femur yields a stature estimate of 168.06 centimeters with an averaged uncertainty of 5.14 centimeters. This estimate lies between the male and female specific results, as intended, and the averaged uncertainty correctly reflects the additional uncertainty introduced by the absence of sex determination. The application handles this calculation automatically when the unknown sex option is selected, without requiring the practitioner to manually combine two separate calculations.

Population-Specific Coefficients and Why They Cannot Be Bypassed

The practical significance of population-specific regression coefficients becomes apparent when the same bone length is run through the models for different ethnic groups. A femur of 45 centimeters produces a stature estimate of 169.93 centimeters for a Thai male, 168.51 centimeters for a European male, 164.85 centimeters for an African male, and 166.64 centimeters for an Asian male, with standard errors ranging from 3.27 centimeters for the European model to 5.06 centimeters for the Thai model. These differences reflect genuine morphological variation in body proportions across populations, specifically in the ratio of long bone length to total stature, and they are not interchangeable.

Using a European coefficient set for a Thai individual does not produce a slightly inaccurate result. It produces a result calibrated to a different biological population, and the direction and magnitude of the resulting error depends on the bone being measured, the specific regression coefficients, and the actual stature of the individual. In a forensic identification context, this kind of systematic error can push a stature estimate just far enough outside the range of a potential match to cause a false exclusion, or just far enough inside the range of a non-match to cause a false inclusion. Neither outcome serves justice.

The higher standard error in the Thai population data, reflecting greater morphological variability within the Mahakkanukrauh et al. (2011) sample relative to the Trotter and Gleser samples, is a feature of the underlying population data, not a limitation of the statistical method. The application propagates it correctly into the uncertainty estimate, ensuring that the precision communicated to the practitioner honestly reflects the precision available from the current scientific literature for that population.

Field and Laboratory Applications

The application runs on any device with a web browser, which in practice means a desktop in a forensic laboratory, a smartphone at a crime scene, or an iPad on an archaeological excavation in a remote field location where carrying specialized hardware is not feasible. The interface is designed around the measurement workflow that practitioners actually use, adding bones one at a time as they become available, recalculating the combined estimate after each entry, and displaying both individual and combined results with their associated uncertainties.

In a field context, this means a practitioner can arrive at a site, begin entering femur measurements as the excavation proceeds, and have a working stature estimate with honest uncertainty bounds before the skeleton has been fully exposed, with the understanding that each additional bone entered will narrow those bounds further. In a laboratory context, it means that a forensic anthropologist working through a disaster victim identification case can process multiple bones from the same individual systematically, watching the estimate converge as measurements accumulate, without switching between formula sets or maintaining parallel calculations on paper.

The diurnal variation in living stature, which can amount to up to 1.5 centimeters of reduction over the course of a day due to intervertebral disc compression and is further influenced by age-related disc degeneration and hydration status, represents a genuine complication in stature estimation from skeletal material. The available regression coefficients were derived from cadaveric measurements or from living subjects measured at specific times under specific conditions, and neither the original studies nor the current application fully account for this variability. This is an honest limitation, and it is one that the development work toward version 3.0 is directly addressing.

Antrometric v3.0: What Is Already in Development

I am working on version 3.0, and I want to be specific about what is driving it, because the motivation comes directly from the limitations that operational use of version 2.1 has made visible.

The largest single gap in the current model is the absence of empirically derived correction factors for diurnal stature variation. The existing regression coefficients assume a static stature, but living human stature is not static. It decreases over the course of a day as intervertebral discs gradually compress under gravitational load, by amounts that vary with age, hydration status, disc health, and the time elapsed since the individual last rose from sleep. An individual measured at 175 centimeters in the morning may measure 173.5 centimeters by evening, and an older individual with degenerative disc changes may show a reduction of up to 2 centimeters over the same interval. When regression coefficients are derived from cadaveric measurements or from living subjects measured at a specific point in the diurnal cycle, the resulting models carry a systematic error component that is currently unquantified.

Version 3.0 addresses this directly through a dataset of X-ray images from living individuals that pairs precise bone length measurements with corresponding stature values recorded at documented times within the diurnal cycle. From this dataset, correction factors that account for the time of measurement, the age of the individual, and the estimated hydration status can be derived and integrated into the regression models, producing a dynamically adjusted stature estimate that is calibrated to the conditions under which the reference measurements were taken. The corrected stature formula incorporates an adjustment term derived from these empirical data,

[math]\text{stature}_{\text{adjusted}} = \text{stature}_{\text{measured}} + \Delta h[/math]

where Δh is a correction factor derived from empirical measurements of diurnal variation as a function of time of day, age, and hydration status, applied to the base regression output. This approach transforms a currently unquantified source of systematic error into a quantified and correctable one.

Additional development directions for version 3.0 include exploration of the influence of nutritional status during developmental years on the long bone to stature ratio, biomechanical factors such as habitual posture and occupational skeletal adaptation, and the potential for machine learning approaches to extract additional predictive signal from the combination of multiple bone measurements beyond what the current weighted mean approach captures. The goal across all of these directions is the same: progressively narrowing the uncertainty bounds on stature estimates from skeletal material, until the gap between what the regression can deliver and what identification work requires becomes small enough to be operationally irrelevant.

Using the Application

Antrometric v2.1 is accessible at antrometric.com, runs without installation in any current web browser, and requires no login or registration. The workflow is straightforward: select the appropriate ethnic group from the available options covering European, African, Asian, and Thai populations; select the sex of the individual or choose unknown; select your preferred unit of measurement in centimeters or inches; add each available bone measurement using the add bone function; and calculate. The application returns individual stature estimates for each bone with their uncertainties, and where multiple bones have been entered, the combined weighted estimate with its reduced combined uncertainty.

The open-source status of the application, released under the MIT License as documented in Rauscher (2025, Technical Documentation of Antrometric 2.1, International Institute of Forensic Expertise, IIFE), means that the code is available for examination, modification, and extension, and that any forensic anthropologist, statistician, or software developer who identifies an improvement to the methodology or the coefficient database is welcome to contribute it. The scientific foundation of stature estimation depends on the accuracy of the underlying population data, and that data continues to grow as new studies extend the coefficient database to additional populations and refine the estimates for those already covered.

What the Numbers Actually Mean

I want to close with something that the technical documentation of any analytical tool tends to obscure: the purpose of the calculation. The number that Antrometric returns is a stature estimate, not a stature measurement. It carries an uncertainty, and that uncertainty is not a defect of the method, it is the method’s honest characterization of what the science can and cannot tell us from the available skeletal material. Using the weighted mean approach with multiple bones reduces that uncertainty. Using population-specific coefficients reduces systematic bias. Using the correct sex-specific model, or the correctly averaged unknown-sex model, ensures that the estimate is calibrated to the right biological reference population. Each of these choices improves the estimate in a specific and documentable way.

What the estimate then does is give the investigation somewhere to look. A stature estimate of 167 centimeters with a combined uncertainty of 3.5 centimeters provides a range of 163.5 to 170.5 centimeters within which the true stature of the individual almost certainly falls. A missing person recorded at 165 centimeters does not fall outside that range. A missing person recorded at 182 centimeters does. This is the forensic utility of the calculation: not certainty, which skeletal remains cannot provide, but a defensible and statistically grounded constraint on the universe of possible identities, one narrow enough to be useful and honest enough to hold up in court.

That is what Antrometric is built for. Version 3.0 will be built for something more precise. The work continues.

References

• Feldesman, M. R., Kleckner, J. G., and Lundy, J. K. (1990). The femur/stature ratio and estimates of stature in mid and late-Pleistocene fossil hominids. Journal of Forensic Sciences, 35(2), 431-446.
• Mahakkanukrauh, P., Khanpetch, P., Prasitwattanaseree, S., Vichairat, K., and Troy Case, D. (2011). Stature estimation from long bone lengths in a Thai population. Forensic Science International, 210(1-3), 279.e1-279.e7.
• Rauscher, G. A. (2025). Technical documentation of Antrometric 2.1. International Institute of Forensic Expertise (IIFE).
• Trotter, M., and Gleser, G. C. (1952). Estimation of stature from long bones of American Whites and Negroes. American Journal of Physical Anthropology, 10(4), 463-514.
• Trotter, M., and Gleser, G. C. (1958). A re-evaluation of estimation of stature based on measurements of stature taken during life and of long bones after death. American Journal of Physical Anthropology, 16(1), 79-123.
• White, A. A., and Panjabi, M. M. (1990). Clinical biomechanics of the spine (2nd ed.). J.B. Lippincott.
• Bogin, B. (2001). The growth of humanity. Wiley-Liss.
• Ruff, C. B. (2002). Variation in human body size and shape. Annual Review of Anthropology, 31, 211-232.
• Steckel, R. H. (1995). Stature and the standard of living. Journal of Economic Literature, 33(4), 1903-1940.