Correction updates Blackbelly sheep weight prediction paper
Bottom line
A correction notice has been issued for the March 5, 2026, BMC Veterinary Research paper on using XGBoost and Random Forest models to predict live weight in Blackbelly sheep from biometric measurements. The notice says errors introduced during proofing were not reflected in the published version, leading to mixed figure ordering and mismatched captions. The underlying study compared the two machine-learning approaches in 120 clinically healthy Blackbelly lambs, ages 6 to 8 months, raised in humid tropical conditions in Tabasco, Mexico, and reported that Random Forest delivered more stable test-set performance than XGBoost for live-weight estimation. (link.springer.com)
Why it matters: For veterinary professionals, this is less about a change in the study’s headline conclusion and more about the reliability of the published record. Papers on biometric weight prediction can inform flock management, growth monitoring, selection decisions, and field-friendly alternatives where scales are limited, so figure and caption accuracy matters if clinicians, researchers, or advisors are interpreting variable importance, model performance, or practical takeaways. In the original article, chest circumference was the most influential predictor in both models, and the authors positioned these tools as decision-support options for small-scale sheep systems. (link.springer.com)
What to watch: Watch for the corrected version to propagate across journal and indexing platforms, and for whether future validation studies in larger, breed-diverse sheep populations support the original finding that Random Forest generalizes better than XGBoost in this setting. (link.springer.com)
Key facts
- Article type
- Correction notice
- Journal
- BMC Veterinary Research
- Original publication date
- March 5, 2026
- Study species
- Blackbelly sheep
- Sample size
- 120 clinically healthy lambs
- Age range
- 6 to 8 months
- Study location
- Tabasco, Mexico
- Issue corrected
- Proofing errors caused mixed figure order and mismatched captions
- Main finding
- Random Forest performed more stably than XGBoost on the test set
A correction has been published for a 2026 BMC Veterinary Research paper examining whether XGBoost or Random Forest can better predict live weight in Blackbelly sheep using simple biometric measurements. According to the correction summary provided in the source material, the issue was not a change to the study question or dataset, but production-related errors: the proof corrections were not reflected in the published version, resulting in mixed figure order and mismatched captions. The original article itself was published March 5, 2026, with the version of record dated April 13, 2026. (link.springer.com)
The original study sits within a growing body of livestock research trying to estimate body weight without relying on scales alone. In sheep production, body weight is a core management variable tied to feeding, breeding, health assessment, and productivity, and prior literature has framed morphometric and biometric measurements as a practical substitute when weighing infrastructure is limited. A broader review and related sheep studies show that machine-learning models are increasingly being tested against traditional regression methods for this purpose. (link.springer.com)
In this case, the authors analyzed 120 Blackbelly lambs, split evenly by sex, from a local breeding farm in Tabasco, Mexico. The lambs were clinically healthy, 6 to 8 months old, and evaluated using body weight plus measurements including heart circumference, cross-body length, abdominal circumference, body length, withers height, rump height, and hip width. The paper’s central finding was that XGBoost fit the training data more tightly, but Random Forest held up better on the test set, suggesting stronger generalization. Test-set performance favored Random Forest, with an R² of 0.873 versus 0.813 for XGBoost, alongside lower RMSE and MAE. (link.springer.com)
That conclusion also fits with some of the comparative literature the authors cited and with newer work in the field, though the broader evidence base remains mixed by breed, dataset size, and feature selection. The Blackbelly paper notes that chest circumference carried the greatest importance in both models, a biologically plausible result given its relationship to skeletal development and body mass. Other recent sheep-weight prediction studies have likewise emphasized the practical value of morphometric measures, while newer work is also moving toward explainable AI frameworks and, in some cases, image-based approaches. (link.springer.com)
I did not find a separate press release or formal expert reaction specifically addressing this correction. What the surrounding literature does show is a consistent industry and research interest in low-cost, field-usable weight estimation tools for small ruminants, especially in settings where repeated weighing can add labor, stress, or cost. One recent explainable machine-learning study in sheep explicitly framed better body-weight prediction as a way to improve feeding decisions, animal selection, and health attention, while reducing repeated handling. (sciencedirect.com)
Why it matters: For veterinary professionals, corrections like this are a reminder that even when a paper’s core result stands, presentation errors can affect interpretation. In a methods-heavy study, swapped figure order or caption mismatches can create confusion about which model performed best, which variables mattered most, or how the authors want readers to interpret overfitting versus generalization. That matters for veterinarians involved in production medicine, research translation, extension work, or flock consulting, because these tools may eventually support growth monitoring, nutrition planning, and breeding decisions in resource-limited systems. (link.springer.com)
There’s also a broader signal here about the trajectory of veterinary and animal science research. Machine learning for body-weight estimation is moving from proof-of-concept toward more practical deployment, but many studies, including this one, still rely on relatively small, breed-specific datasets. The authors themselves said larger datasets and validation across additional breeds will be needed to strengthen reliability and generalizability, which is especially important before clinicians or producers treat these models as interchangeable with direct measurement. (link.springer.com)
What to watch: The next step is straightforward but important: readers should watch for the corrected article version to be fully reflected in journal databases, and for future sheep studies to test whether Random Forest’s apparent edge in Blackbelly lambs holds up in larger populations, other breeds, and image-based or explainable-model workflows. (link.springer.com)