README

Stop wasting days manually assembling complex variance-covariance matrices and resolving selection weights. selection.index is the definitive, production-ready R package built specifically for agricultural statisticians and quantitative geneticists to instantly calculate optimal multi-trait selection indices. Whether you are running classical phenotypic field trials or building advanced genomic evaluations, this package provides a fully integrated mathematical engine that directly maximizes genetic advance and substantially accelerates your breeding pipeline.

Installation

install.packages("selection.index")

# install.packages("devtools")
devtools::install_github("zankrut20/selection.index")

Quick Start: Optimizing Phenotypic Selection

The package includes several pre-configured structural datasets (seldata, maize_pheno, maize_geno) to test your models. Here, we demonstrate a fundamental phenotypic index evaluating multiple field traits simultaneously.

Before combining traits into a unified index, breeders must cleanly partition the total phenotypic variation. We first use the variance-covariance functions (phen_varcov() and gen_varcov()) to extract the precise phenotypic and genotypic matrices directly from your replicated trial data.

# Using the built-in 'seldata' dataset (Replication in Col 1, Genotype in Col 2)
p_mat <- phen_varcov(data = seldata[, 3:9], genotypes = seldata[, 2], replication = seldata[, 1])
g_mat <- gen_varcov(data = seldata[, 3:9], genotypes = seldata[, 2], replication = seldata[, 1])

Next, we establish strict economic objectives for our evaluated traits. The weight_mat() function rapidly converts raw economic weight arrays into the formalized mathematical format required for the core index calculation.

# Convert the built-in 'weight' configuration matrix
w_mat <- weight_mat(weight)

With our covariance structures and economic priorities firmly defined, we must evaluate the baseline potential of our population. We calculate the expected genetic gain for our primary target trait (e.g., yield) using the gen_advance() function.

# Calculate expected genetic advance for the primary trait
base_advance <- gen_advance(phen_mat = p_mat[1, 1], gen_mat = g_mat[1, 1], weight_mat = w_mat[1, 1])
base_advance
#>        [,1]
#> [1,] 1.2009

Finally, to make definitive selection decisions, we dynamically evaluate trait combinations. The lpsi() function constructs robust combinatorial selection indices and immediately ranks them by predicted relative efficiency, allowing you to confidently select the mathematically superior combination.

# Calculate and rank combinatorial selection indices
optimal_index <- lpsi(
  ncomb = 1, pmat = p_mat, gmat = g_mat,
  wmat = weight[, 2:3], wcol = 1, GAY = base_advance
)
head(optimal_index)
#>   ID    b.1     GA      PRE Delta_G    rHI    hI2 Rank
#> 1  1 0.6316 2.8125 234.2021  2.8125 0.4825 0.2697    1
#> 2  2 0.2396 1.3036 108.5481  1.3036 0.2237 0.2242    4
#> 3  3 1.4869 2.1526 179.2462  2.1526 0.3693 0.2690    2
#> 4  4 0.4507 0.9084  75.6402  0.9084 0.1558 0.2549    6
#> 5  5 0.3164 1.6236 135.1988  1.6236 0.2786 0.1553    3
#> 6  6 1.4499 1.0292  85.7009  1.0292 0.1766 0.1471    5

(For advanced modeling involving marker-assisted indices, multistage selection, or genomic predictions, please load the included maize_pheno and maize_geno datasets and consult our pkgdown vignettes.)

Citation

We rely directly on academic citations to maintain and justify the continued development of this software ecosystem. If you use selection.index in a publication, breeding pipeline, or research project, it is imperative that you cite it:

citation("selection.index")

Support & Consulting

Direct-to-Client Consulting

Are you scaling up a commercial breeding operation or standardizing a massive academic quantitative pipeline? I offer dedicated, expert consulting to researchers tailored for custom selection index development, stochastic simulation modeling, and full software pipeline integration.

Note: I operate strictly direct-to-client. There are absolutely no middlemen, generic agencies, or intermediaries involved. You work directly with the package creator.

Tip Jar

If this package has saved you substantial time, simplified your data pipeline, or helped you publish faster, please consider supporting its independent development! Monetary contributions directly fund repository maintenance and the addition of modern quantitative genetic models.