| 1 | test_that("on input of sufficient size, β/||β|| is estimated accurately enough", |
| 2 | { |
| 3 | n <- 100000 |
| 4 | d <- 2 |
| 5 | K <- 2 |
| 6 | p <- 1/2 |
| 7 | |
| 8 | βs_ref <- array( c(1,0,0,1 , 1,-2,3,1), dim=c(d,K,2) ) |
| 9 | for (i in 1:(dim(βs_ref)[3])) |
| 10 | { |
| 11 | μ_ref <- normalize(βs_ref[,,i]) |
| 12 | for (link in c("logit","probit")) |
| 13 | { |
| 14 | cat("\n\n",link," :\n",sep="") |
| 15 | |
| 16 | io <- generateSampleIO(n, p, βs_ref[,,i], rep(0,K), link) |
| 17 | μ <- computeMu(io$X, io$Y, list(K=K)) |
| 18 | μ_aligned <- alignMatrices(list(μ), ref=μ_ref, ls_mode="exact")[[1]] |
| 19 | |
| 20 | #Some traces: 0 is not well estimated, but others are OK |
| 21 | cat("Reference normalized matrix:\n") |
| 22 | print(μ_ref) |
| 23 | cat("Estimated normalized matrix:\n") |
| 24 | print(μ_aligned) |
| 25 | cat("Difference norm (Matrix norm ||.||_1, max. abs. sum on a column)\n") |
| 26 | diff_norm <- norm(μ_ref - μ_aligned) |
| 27 | cat(diff_norm,"\n") |
| 28 | |
| 29 | # NOTE: 0.5 is loose threshold, but values around 0.3 are expected... |
| 30 | expect_lt( diff_norm, 0.5 ) |
| 31 | } |
| 32 | } |
| 33 | }) |