`fit1_pvl` uses several functions in the package qtl2pleio to fit the linear mixed effects model for a single d-tuple of markers. Creation of `fit1_pvl` - from code that originally resided in `scan_pvl`, enabled parallelization via the `parallel` R package.
fit1_pvl(indices, start_snp, probs, addcovar, inv_S, S, pheno)
| indices | a vector of indices for extracting elements of `probs` array |
|---|---|
| start_snp | an integer to specify the index of the marker where the scan - in call to scan_pvl - starts. This argument is needed because `mytab` has only relative indices (relative to the `start_snp` marker) |
| probs | founder allele probabilities array |
| addcovar | additive covariates matrix |
| inv_S | inverse covariance matrix for the vectorized phenotype |
| S | covariance matrix for the vectorized phenotype, ie, the inverse of inv_S. By making this a function input, we avoid inverting the matrix many many times. |
| pheno | a n by d phenotypes matrix |
a number, the log-likelihood for the specified model
n <- 50 pheno <- matrix(rnorm(2 * n), ncol = 2) Vg <- diag(2) Ve <- diag(2) Sigma <- calc_Sigma(Vg, Ve, diag(n)) Sigma_inv <- solve(Sigma) probs <- array(dim = c(n, 2, 5)) probs[ , 1, ] <- rbinom(n * 5, size = 1, prob = 0.2) probs[ , 2, ] <- 1 - probs[ , 1, ] mytab <- prep_mytab(d_size = 2, n_snp = 5) fit1_pvl(mytab[1, ], start_snp = 1, probs = probs, addcovar = NULL, inv_S = Sigma_inv, S = Sigma, pheno = pheno )#> [1] -151.261