rrBLUP_ML0#

pybrops.model.gmod.rrBLUPModel0.rrBLUP_ML0(y, Z, varlb=1e-05, varub=100000.0, gsatol=1e-08, gsmaxiter=1000)[source]#

Ridge regression BLUP for the simple model:

y = Zu + e

Where:

- ``y`` are observations.
- ``Z`` is a design matrix for genetic markers.
- ``u`` are marker effects which follow the distribution ``MVN(0, varU * I)``.
- ``e`` are errors which follow the distribution ``MVN(0, varE * I)``.

Uses the EMMA formulation to solve for varE and varU. Uses the Nelder-Mead method to optimize for variance components. Marker effects are estimated using the Gauss-Seidel method.

Parameters:

y (numpy.ndarray) – A vector of observations of shape (nobs,). If not mean centered, will be centered around zero.
Z (numpy.ndarray) – A genotype matrix of shape (nobs,nmkr).
varlb (Real) – Lower bound permitted for variance component estimation. Must be non-negative.
varub (Real) – Upper bound permitted for variance component estimation. Must be non-negative and greater than varlb.
gsatol (Real) – Absolute tolerance for the Gauss-Seidel method. Iterate until the sum of absolute differences between successive iterations is less than this value or maxiter is reached. Must be non-negative.
gsmaxiter (Integral) – Maximum number of iterations for the Gauss-Seidel method. Must be non-negative.

Returns:

out – A dictionary of output values.

Return type:

dict