Age-dependent morphometric variation and internally validated live-weight prediction in phenotypically identified Waila sheep of northern Cameroon
Year: 2026
Type: preprint
Abstract: Live-weight equations require population-specific validation for use where scales are unavailable. We analysed 287 phenotypically identified Waila sheep (177 males, 110 females) from four departments of northern Cameroon. Sex differences were modelled across four age classes (≤ 12, > 12–24, > 24–36 and > 36–60 months), including sex-by-age interactions and department adjustment. Morphometric structure was assessed using principal component analysis and Ward clustering. Predictor selection was repeated within validation folds; fixed linear and machine-learning models used identical 10-fold partitions repeated ten times. Sex differences in live weight and four principal measurements depended on age (global interaction P ≤ 0.011). Adjusted male–female weight contrasts were negligible at ≤ 12 months and ranged from + 9.21 to + 20.91 kg thereafter. One general-size component explained 69.3% of morphometric variance. Clustering this component favoured six groups by silhouette width within the tested range of two to six, partitioning an age- and sex-associated size continuum without establishing distinct morphotypes. Neck circumference had the strongest pooled correlation with weight (r = 0.800) and lower cross-validated root mean squared error (RMSE) than chest girth (10.08 versus 10.70 kg), but no consistent within-sex advantage. The three-measurement equation combining neck circumference, scapulo-ischial length and chest girth achieved repeated cross-validation R² = 0.780 and RMSE = 7.79 kg. The six-measurement stepwise equation achieved nested cross-validation R² = 0.790 and RMSE = 7.63 kg. Fixed machine-learning configurations showed no consistent improvement across error metrics. These internally validated equations support approximate weight estimation but require independent-flock validation before decision-specific use.
Source: Research Square
Institution: Université de Dschang
Language: English
Subfield: Genetics
Domain: Life Sciences
Open Access status: green
OpenAlex ID: W7225778452