In andrewpbray/lmmoptim: Optimization routine for linear mixed models

Todo

1. Write general pseudocode.
2. Lay out proof that bounds will go to true value (sandwich, continuity, how general can this proof be made?).
3. Re-run figures and verify the second mode.

library(microbenchmark)
x <- matrix(rnorm(1e5), ncol = 5, byrow = TRUE)
microbenchmark(x[x[, 1] + x[, 2] < x[, 3], 5] <- 0)

y <- matrix(c(rnorm(1e5), rep(NA, 1e6)), ncol = 5, byrow = TRUE)
microbenchmark(y[y[, 1] + y[, 2] < y[, 3], 5] <- 0)

SET epsilon create data structure

WHILE there are still active boxes Bound new boxes IF Bounds

• addprior: sensible defaults for the alpha/beta params?

andrewpbray/lmmoptim documentation built on May 10, 2019, 11:10 a.m.