- Mixed models are important statistical tools in animal breeding and quantitative genetics because they help researchers separate genetic influences from environmental effects when evaluating animal performance. Traits such as milk production, growth rate, fertility, feed efficiency, disease resistance, and meat quality are influenced by many factors. These include an animal’s genetic makeup, nutrition, management, age, health, climate, and housing conditions. Mixed models provide a systematic way to account for these different sources of variation and make more reliable estimates of genetic merit.
- A mixed model is a statistical model that includes both fixed effects and random effects. Fixed effects represent influences whose categories or specific values are of direct interest or are used to adjust the data. Examples include sex, herd, season, year of birth, feeding system, and management group. Random effects represent sources of variation associated with a population of possible effects, such as the genetic differences among animals or the permanent environmental influences experienced by individual animals. By including both types of effects in one model, researchers can analyze complex biological data more accurately than with many simple statistical methods.
- The general form of a linear mixed model can be written as y=Xb+Zu+e, where y represents the observed records, X is the design matrix relating observations to fixed effects, b represents the fixed-effect solutions, Z is the design matrix relating observations to random effects, u represents the random effects, and e represents the residual errors. The residuals account for variation that remains unexplained by the other effects included in the model. In animal breeding, the random effects often include the additive genetic effects of animals, which are particularly important for estimating breeding values.
- Fixed effects are necessary because animals are frequently raised under different conditions. For example, a cow producing milk in a well-managed herd with high-quality feed may produce more milk than a genetically similar cow raised under less favorable conditions. If these environmental differences are ignored, the first cow could be incorrectly considered genetically superior. Including herd, year, season, and other relevant management factors helps account for systematic environmental differences. This allows genetic evaluations to focus more closely on inherited differences rather than advantages or disadvantages caused by the production environment.
- Random genetic effects are central to mixed models in quantitative genetics. Animals differ in their inherited ability to express traits, and these differences can be estimated using information from their own records and the records of genetically related animals. An animal’s additive genetic effect represents the part of its genetic merit associated with the average effects of alleles inherited from its parents. These additive genetic effects form the basis of estimated breeding values, which are used to identify animals that are likely to produce genetically superior offspring.
- One of the most important applications of mixed models is the estimation of breeding values. A breeding value predicts the genetic contribution an animal is expected to pass to its offspring for a particular trait. For example, a dairy cow may have an estimated breeding value for milk yield, while a beef bull may have estimated breeding values for growth rate, carcass quality, and feed efficiency. Mixed models combine performance records, information about relationships among animals, and relevant environmental factors to produce these estimates. Animals with higher breeding values for desirable traits can then be selected as parents for future generations.
- The animal model is one of the most widely used mixed models in animal breeding. It uses pedigree information or genomic relationship information to account for genetic similarities among animals. The model recognizes that relatives share genes and that their records provide useful information about one another’s genetic merit. An animal with few or no individual performance records may still receive a useful genetic evaluation if it has parents, siblings, offspring, or other relatives with informative records. This is especially valuable for traits that are expensive to measure, expressed late in life, or limited to one sex, such as milk production in dairy cattle.
- Best linear unbiased prediction, commonly known as BLUP, is closely associated with mixed models in animal breeding. BLUP is a method for predicting random effects, including breeding values, while accounting for fixed effects in the statistical model. It is widely used because it can combine information from animals with different numbers of records, different levels of relatedness, and different production environments. It also accounts for the fact that genetic predictions should be influenced by the reliability and amount of available information. For example, an animal with many informative relatives and repeated records may have a more reliable breeding value than an animal with only one observation.
- A major advantage of mixed models is their ability to account for unbalanced data. In livestock production, it is uncommon for every animal to have the same number of observations or to be raised under identical conditions. Some cows may have several lactation records, while others may have only one. Some animals may be tested in multiple environments, while others may be evaluated in a single herd. Traditional statistical methods can become difficult to apply appropriately when data are unbalanced or observations are not independent. Mixed models are designed to handle many such situations, provided the model assumptions and data structure are suitable.
- Repeated measurements are another important reason to use mixed models. Many animal traits are recorded several times during an animal’s life. Dairy cows, for instance, may have milk production measured across multiple lactations, while growing animals may be weighed at different ages. Measurements taken from the same animal are generally correlated because they share genetic and environmental influences. A mixed model can include an animal-level random effect to account for persistent differences among animals. More advanced models can also represent changing genetic effects over time, such as random regression models used to analyze milk production across the lactation curve.
- Mixed models are also useful for estimating variance components, which describe how much variation in a trait is associated with different sources. In quantitative genetics, researchers often want to estimate additive genetic variance, permanent environmental variance, and residual variance. These estimates help explain why animals differ and determine how much of the observed variation may respond to selection. Under an appropriate model, heritability can be estimated as the ratio of additive genetic variance to total phenotypic variance. Heritability provides an indication of how strongly observed differences in a population are associated with additive genetic differences, although its value depends on the population, trait, environment, and statistical model used.
- For example, suppose researchers want to improve body weight in a population of beef cattle. They collect weights from calves born in different years, raised in different herds, and managed under different feeding systems. A mixed model can include year, herd, sex, and age at weighing as fixed effects, while including the animals’ additive genetic effects as random effects. If repeated observations are available, the model can also account for permanent environmental effects. The resulting genetic evaluations help distinguish calves that are heavier because of favorable management from those whose weight reflects superior inherited potential. This supports more informed selection decisions.
- In dairy cattle breeding, mixed models are commonly used to evaluate milk yield, fat percentage, protein percentage, fertility, udder health, and longevity. Environmental conditions can vary substantially between herds and across years, making adjustment essential for fair genetic comparisons. Genetic evaluations may combine records from many cows, their ancestors, and their descendants. Depending on the evaluation system, genomic information may also be included through genomic relationship matrices or other genomic prediction approaches. These methods allow breeders to evaluate animals earlier in life and improve the speed and accuracy of genetic selection.
- Mixed models are not limited to production traits. They can also be applied to fertility, disease resistance, survival, behavior, and other traits important for animal welfare and sustainable livestock production. However, the statistical model must match the nature of the data. For continuous traits, a linear mixed model may be appropriate. For binary traits, such as whether an animal developed a particular disease, generalized linear mixed models may be more suitable. These models use an appropriate link function and probability distribution to represent the relationship between explanatory variables and the outcome. Survival models and other specialized approaches may be needed for time-to-event traits or more complex outcomes.
- The development of genomic technologies has further expanded the role of mixed models in animal breeding. Genomic information can improve the estimation of relationships among animals and help predict genetic merit even when individual performance records are unavailable. Genomic best linear unbiased prediction, or GBLUP, uses genomic relationships to estimate breeding values. Other methods, such as single-step genomic BLUP, combine pedigree, genomic, and phenotypic information in a unified evaluation framework. These approaches are particularly valuable for young animals that have not yet produced offspring or expressed traits that can be measured directly.
- Despite their advantages, mixed models require careful construction and interpretation. The choice of fixed effects, random effects, covariance structures, and variance assumptions can influence the results. Missing records, incorrect pedigree information, measurement errors, and inadequate sample sizes may reduce the quality of genetic evaluations. If important environmental factors are omitted, genetic estimates may be biased. Similarly, an overly complicated model may be difficult to estimate or may not provide meaningful improvements over a simpler model. Researchers must therefore understand the biological process behind the data and select a model that appropriately reflects the breeding population and research objectives.
- Another important consideration is the distinction between prediction and causation. Mixed models can estimate genetic effects and account for observed environmental differences, but their results do not automatically prove that a particular factor causes a trait to change. The reliability of the conclusions depends on the quality of the data, the assumptions of the model, and the suitability of the statistical design. Genetic evaluations should also be interpreted in relation to the population and evaluation system in which they were produced. Breeding values from different populations or evaluation programs may not be directly comparable unless appropriate genetic and environmental connections have been established.
- Modern statistical software has made mixed-model analysis more accessible to animal geneticists and breeders. Specialized programs and statistical packages can estimate variance components, predict breeding values, analyze repeated measurements, and incorporate pedigree or genomic relationships. However, software cannot compensate for poor data quality or an unsuitable model. Researchers need to verify the structure of the dataset, inspect the assumptions, assess the precision of estimates, and communicate the results clearly to breeders and livestock managers.
- Mixed models have become fundamental to modern animal breeding because they provide a flexible framework for analyzing complex biological data. By combining fixed environmental effects with random genetic and other effects, they improve the fairness of animal comparisons, support the estimation of breeding values, and help researchers understand the sources of variation in economically and biologically important traits. Their integration with pedigree records, repeated measurements, and genomic information has strengthened genetic evaluation systems across livestock species. As animal breeding increasingly emphasizes productivity, health, fertility, adaptability, and sustainability, mixed models will continue to play an essential role in guiding reliable genetic improvement and more informed breeding decisions.
