Refactor mkinfit, infrastructure work

mkinfit objects now include an ll() function to calculate the
log-likelihood. Part of the code was refactored, hopefully making it
easier to read and maintain. IRLS is currently the default algorithm for
the error model "obs", for no particular reason. This may be subject
to change when I get around to investigate.

Slow tests are now in a separate subdirectory and will probably
only be run by my own Makefile target.

Formatting of test logs is improved.

Roundtripping error model parameters works with a precision of 10% when
we use lots of replicates in the synthetic data (see slow tests). This
is not new in this commit, but as I think it is reasonable this
closes #7.

2 files changed, 359 insertions, 0 deletions

diff --git a/tests/testthat/slow/test_parent_only.R b/tests/testthat/slow/test_parent_only.R
new file mode 100644
index 00000000..7521e145
--- /dev/null
+++ b/tests/testthat/slow/test_parent_only.R
@@ -0,0 +1,218 @@
+# Copyright (C) 2015,2018 Johannes Ranke
+# Contact: jranke@uni-bremen.de
+
+# This file is part of the R package mkin
+
+# mkin is free software: you can redistribute it and/or modify it under the
+# terms of the GNU General Public License as published by the Free Software
+# Foundation, either version 3 of the License, or (at your option) any later
+# version.
+
+# This program is distributed in the hope that it will be useful, but WITHOUT
+# ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
+# FOR A PARTICULAR PURPOSE.  See the GNU General Public License for more
+# details.
+
+# You should have received a copy of the GNU General Public License along with
+# this program. If not, see <http://www.gnu.org/licenses/>
+
+context("Fitting of parent only models")
+
+calc_dev.percent <- function(fitlist, reference, endpoints = TRUE, round_results = NULL) {
+  dev.percent <- list()
+  for (i in 1:length(fitlist)) {
+    fit <- fitlist[[i]]
+    if (endpoints) {
+      results <- c(fit$bparms.optim,
+                   endpoints(fit)$distimes$DT50,
+                   endpoints(fit)$distimes$DT90)
+    } else {
+      results <- fit$bparms.optim
+    }
+    if (!missing(round_results)) results <- round(results, round_results)
+    dev.percent[[i]] <- abs(100 * ((reference - results)/reference))
+  }
+  return(dev.percent)
+}
+
+SFO <- mkinmod(parent = list(type = "SFO"))
+FOMC <- mkinmod(parent = list(type = "FOMC"))
+DFOP <- mkinmod(parent = list(type = "DFOP"))
+HS <- mkinmod(parent = list(type = "HS"))
+SFORB <- mkinmod(parent = list(type = "SFORB"))
+
+test_that("Fits for FOCUS A deviate less than 0.1% from median of values from FOCUS report", {
+  fit.A.SFO <- list(mkinfit("SFO", FOCUS_2006_A, quiet = TRUE))
+
+  median.A.SFO <- as.numeric(lapply(subset(FOCUS_2006_SFO_ref_A_to_F,
+                                        dataset == "A",
+                                        c(M0, k, DT50, DT90)), "median"))
+
+  dev.percent.A.SFO <- calc_dev.percent(fit.A.SFO, median.A.SFO)
+  expect_equivalent(dev.percent.A.SFO[[1]] < 0.1, rep(TRUE, 4))
+
+  # Fitting FOCUS A with FOMC is possible, but the correlation between
+  # alpha and beta, when obtained, is 1.0000, and the fit does not
+  # always converge using the Port algorithm (platform dependent), so
+  # we need to suppress a potential warning
+  suppressWarnings(fit.A.FOMC <- try(list(mkinfit("FOMC", FOCUS_2006_A, quiet = TRUE))))
+
+  if (!inherits(fit.A.FOMC, "try-error")) {
+
+    median.A.FOMC <- as.numeric(lapply(subset(FOCUS_2006_FOMC_ref_A_to_F,
+                                          dataset == "A",
+                                          c(M0, alpha, beta, DT50, DT90)), "median"))
+
+    dev.percent.A.FOMC <- calc_dev.percent(fit.A.FOMC, median.A.FOMC)
+    # alpha and are beta ill-determined, do not compare those
+    expect_equivalent(dev.percent.A.FOMC[[1]][c(1, 4, 5)] < 0.1, rep(TRUE, 3))
+  }
+
+  fit.A.DFOP <- list(mkinfit("DFOP", FOCUS_2006_A, quiet = TRUE))
+
+  median.A.DFOP <- as.numeric(lapply(subset(FOCUS_2006_DFOP_ref_A_to_B,
+                                        dataset == "A",
+                                        c(M0, k1, k2, f, DT50, DT90)), "median"))
+
+  dev.percent.A.DFOP <- calc_dev.percent(fit.A.DFOP, median.A.DFOP)
+  #expect_equivalent(dev.percent.A.DFOP[[1]] < 0.1, rep(TRUE, 6)) # g/f is ill-determined
+  expect_equivalent(dev.percent.A.DFOP[[1]][c(1, 2, 3, 5, 6)] < 0.1, rep(TRUE, 5))
+
+  fit.A.HS <- list(mkinfit("HS", FOCUS_2006_A, quiet = TRUE))
+
+  median.A.HS <- as.numeric(lapply(subset(FOCUS_2006_HS_ref_A_to_F,
+                                        dataset == "A",
+                                        c(M0, k1, k2, tb, DT50, DT90)), "median"))
+
+  dev.percent.A.HS <- calc_dev.percent(fit.A.HS, median.A.HS)
+  expect_equivalent(dev.percent.A.HS[[1]] < 0.1, rep(TRUE, 6))
+})
+
+test_that("Fits for FOCUS B deviate less than 0.1% from median of values from FOCUS report", {
+  skip_on_cran()
+  fit.B.SFO <- list(mkinfit("SFO", FOCUS_2006_B, quiet = TRUE))
+
+  median.B.SFO <- as.numeric(lapply(subset(FOCUS_2006_SFO_ref_A_to_F,
+                                        dataset == "B",
+                                        c(M0, k, DT50, DT90)), "median"))
+
+  dev.percent.B.SFO <- calc_dev.percent(fit.B.SFO, median.B.SFO)
+  expect_equivalent(dev.percent.B.SFO[[1]] < 0.1, rep(TRUE, 4))
+
+  fit.B.FOMC <- list(mkinfit("FOMC", FOCUS_2006_B, quiet = TRUE))
+
+  median.B.FOMC <- as.numeric(lapply(subset(FOCUS_2006_FOMC_ref_A_to_F,
+                                        dataset == "B",
+                                        c(M0, alpha, beta, DT50, DT90)), "median"))
+
+  dev.percent.B.FOMC <- calc_dev.percent(fit.B.FOMC, median.B.FOMC)
+  expect_equivalent(dev.percent.B.FOMC[[1]] < 0.1, rep(TRUE, 5))
+
+  fit.B.DFOP <- list(mkinfit("DFOP", FOCUS_2006_B, quiet = TRUE))
+
+  median.B.DFOP <- as.numeric(lapply(subset(FOCUS_2006_DFOP_ref_A_to_B,
+                                        dataset == "B",
+                                        c(M0, k1, k2, f, DT50, DT90)), "median"))
+
+  dev.percent.B.DFOP <- calc_dev.percent(fit.B.DFOP, median.B.DFOP)
+  #expect_equivalent(dev.percent.B.DFOP[[1]] < 0.1, rep(TRUE, 6)) # g/f is ill-determined
+  expect_equivalent(dev.percent.B.DFOP[[1]][c(1, 2, 3, 5, 6)] < 0.1, rep(TRUE, 5))
+
+  fit.B.HS <- list(mkinfit("HS", FOCUS_2006_B, quiet = TRUE))
+
+  median.B.HS <- as.numeric(lapply(subset(FOCUS_2006_HS_ref_A_to_F,
+                                        dataset == "B",
+                                        c(M0, k1, k2, tb, DT50, DT90)),
+                                   "median", na.rm = TRUE))
+
+  dev.percent.B.HS <- calc_dev.percent(fit.B.HS, median.B.HS)
+  expect_equivalent(dev.percent.B.HS[[1]] < 0.1, rep(TRUE, 6))
+
+  fit.B.SFORB <- list(mkinfit(SFORB, FOCUS_2006_B, quiet=TRUE))
+  dev.percent.B.SFORB <- calc_dev.percent(fit.B.SFORB, median.B.DFOP)
+  expect_equivalent(dev.percent.B.SFORB[[1]][c(1, 5, 6)] < 0.1, rep(TRUE, 3))
+})
+
+test_that("Fits for FOCUS C deviate less than 0.1% from median of values from FOCUS report", {
+  fit.C.SFO <- list(mkinfit("SFO", FOCUS_2006_C, quiet = TRUE))
+
+  median.C.SFO <- as.numeric(lapply(subset(FOCUS_2006_SFO_ref_A_to_F,
+                                        dataset == "C",
+                                        c(M0, k, DT50, DT90)), "median"))
+
+  dev.percent.C.SFO <- calc_dev.percent(fit.C.SFO, median.C.SFO)
+  expect_equivalent(dev.percent.C.SFO[[1]] < 0.1, rep(TRUE, 4))
+
+  fit.C.FOMC <- list(mkinfit("FOMC", FOCUS_2006_C, quiet = TRUE))
+
+  median.C.FOMC <- as.numeric(lapply(subset(FOCUS_2006_FOMC_ref_A_to_F,
+                                        dataset == "C",
+                                        c(M0, alpha, beta, DT50, DT90)), "median"))
+
+  dev.percent.C.FOMC <- calc_dev.percent(fit.C.FOMC, median.C.FOMC,
+                                         round_results = 2) # Not enough precision in FOCUS results
+  expect_equivalent(dev.percent.C.FOMC[[1]] < 0.1, rep(TRUE, 5))
+
+  fit.C.HS <- list(mkinfit("HS", FOCUS_2006_C, quiet = TRUE))
+
+  median.C.HS <- as.numeric(lapply(subset(FOCUS_2006_HS_ref_A_to_F,
+                                        dataset == "C",
+                                        c(M0, k1, k2, tb, DT50, DT90)), "median"))
+
+  dev.percent.C.HS <- calc_dev.percent(fit.C.HS, median.C.HS, round_results = c(2, 4, 6, 2, 2))
+  # Not enouth precision in k2 available
+  expect_equivalent(dev.percent.C.HS[[1]] < c(0.1, 0.1, 0.3, 0.1, 0.1, 0.1), rep(TRUE, 6))
+})
+
+test_that("SFO fits give approximately (0.001%) equal results with different solution methods", {
+  skip_on_cran()
+  fit.A.SFO.default <- mkinfit("SFO", FOCUS_2006_A, quiet = TRUE)$bparms.optim
+
+  fits.A.SFO <- list()
+  fits.A.SFO[[1]] <- mkinfit(SFO, FOCUS_2006_A, quiet = TRUE)
+  fits.A.SFO[[2]] <- mkinfit(SFO, FOCUS_2006_A, quiet = TRUE, solution_type = "eigen")
+  fits.A.SFO[[3]] <- mkinfit(SFO, FOCUS_2006_A, quiet = TRUE, solution_type = "deSolve")
+
+  dev.percent <- calc_dev.percent(fits.A.SFO, fit.A.SFO.default, endpoints = FALSE)
+  expect_equivalent(dev.percent[[1]] < 0.001, rep(TRUE, 2))
+  expect_equivalent(dev.percent[[2]] < 0.001, rep(TRUE, 2))
+  expect_equivalent(dev.percent[[3]] < 0.001, rep(TRUE, 2))
+})
+
+test_that("FOMC fits give approximately (0.001%) equal results with different solution methods", {
+  skip_on_cran()
+  fit.C.FOMC.default <- mkinfit("FOMC", FOCUS_2006_C, quiet = TRUE)$bparms.optim
+
+  fits.C.FOMC <- list()
+  fits.C.FOMC[[1]] <- mkinfit(FOMC, FOCUS_2006_C, quiet = TRUE)
+  fits.C.FOMC[[2]] <- mkinfit(FOMC, FOCUS_2006_C, quiet = TRUE, solution_type = "deSolve")
+
+  dev.percent <- calc_dev.percent(fits.C.FOMC, fit.C.FOMC.default, endpoints = FALSE)
+  expect_equivalent(dev.percent[[1]] < 0.001, rep(TRUE, 3))
+  expect_equivalent(dev.percent[[2]] < 0.001, rep(TRUE, 3))
+})
+
+test_that("DFOP fits give approximately (0.001%) equal results with different solution methods", {
+  skip_on_cran()
+  fit.C.DFOP.default <- mkinfit("DFOP", FOCUS_2006_C, quiet = TRUE)$bparms.optim
+
+  fits.C.DFOP <- list()
+  fits.C.DFOP[[1]] <- mkinfit(DFOP, FOCUS_2006_C, quiet = TRUE)
+  fits.C.DFOP[[2]] <- mkinfit(DFOP, FOCUS_2006_C, quiet = TRUE, solution_type = "deSolve")
+
+  dev.percent <- calc_dev.percent(fits.C.DFOP, fit.C.DFOP.default, endpoints = FALSE)
+  expect_equivalent(dev.percent[[1]] < 0.001, rep(TRUE, 4))
+  expect_equivalent(dev.percent[[2]] < 0.001, rep(TRUE, 4))
+})
+
+test_that("SFORB fits give approximately (0.002%) equal results with different solution methods", {
+  skip_on_cran()
+  fit.B.SFORB.default <- mkinfit(SFORB, FOCUS_2006_B, quiet=TRUE)$bparms.optim
+
+  fits.B.SFORB <- list()
+  fits.B.SFORB[[1]] <- mkinfit(SFORB, FOCUS_2006_B, quiet=TRUE, solution_type = "eigen")
+  fits.B.SFORB[[2]] <- mkinfit(SFORB, FOCUS_2006_B, quiet=TRUE, solution_type = "deSolve")
+  dev.percent <- calc_dev.percent(fits.B.SFORB, fit.B.SFORB.default, endpoints = FALSE)
+  expect_equivalent(dev.percent[[1]] < 0.001, rep(TRUE, 4))
+  expect_equivalent(dev.percent[[2]] < 0.002, rep(TRUE, 4))
+})
diff --git a/tests/testthat/slow/test_roundtrip_error_parameters.R b/tests/testthat/slow/test_roundtrip_error_parameters.R
new file mode 100644
index 00000000..97510563
--- /dev/null
+++ b/tests/testthat/slow/test_roundtrip_error_parameters.R
@@ -0,0 +1,141 @@
+test_that("Reweighting method 'tc' produces reasonable variance estimates", {
+
+  # Check if we can approximately obtain the parameters and the error model
+  # components that were used in the data generation
+
+  # Parent only
+  DFOP <- mkinmod(parent = mkinsub("DFOP"))
+  sampling_times = c(0, 1, 3, 7, 14, 28, 60, 90, 120)
+  parms_DFOP <- c(k1 = 0.2, k2 = 0.02, g = 0.5)
+  parms_DFOP_optim <- c(parent_0 = 100, parms_DFOP)
+
+  d_DFOP <- mkinpredict(DFOP,
+     parms_DFOP, c(parent = 100),
+     sampling_times)
+  d_2_10 <- add_err(d_DFOP,
+    sdfunc = function(x) sigma_twocomp(x, 0.5, 0.07),
+    n = 10, reps = 2, digits = 5, LOD = -Inf, seed = 123456)
+  d_100_1 <- add_err(d_DFOP,
+    sdfunc = function(x) sigma_twocomp(x, 0.5, 0.07),
+    n = 1, reps = 100, digits = 5, LOD = -Inf, seed = 123456)
+
+  # Per default (on my box where I set NOT_CRAN) use all cores minus one
+  if (identical(Sys.getenv("NOT_CRAN"), "true")) {
+    n_cores <- parallel::detectCores() - 1
+  } else {
+    n_cores <- 1
+  }
+
+  # We are only allowed one core on travis, but they also set NOT_CRAN=true
+  if (Sys.getenv("TRAVIS") != "") n_cores = 1
+
+  # On Windows we would need to make a cluster first
+  if (Sys.info()["sysname"] == "Windows") n_cores = 1
+
+  # Unweighted fits
+  f_2_10 <- mmkin("DFOP", d_2_10, error_model = "const", quiet = TRUE,
+    cores = n_cores)
+  parms_2_10 <- apply(sapply(f_2_10, function(x) x$bparms.optim), 1, mean)
+  parm_errors_2_10 <- (parms_2_10 - parms_DFOP_optim) / parms_DFOP_optim
+  expect_true(all(abs(parm_errors_2_10) < 0.12))
+
+  f_2_10_tc <- mmkin("DFOP", d_2_10, error_model = "tc", quiet = TRUE,
+    cores = n_cores)
+  parms_2_10_tc <- apply(sapply(f_2_10_tc, function(x) x$bparms.optim), 1, mean)
+  parm_errors_2_10_tc <- (parms_2_10_tc - parms_DFOP_optim) / parms_DFOP_optim
+  expect_true(all(abs(parm_errors_2_10_tc) < 0.05))
+
+  tcf_2_10_tc <- apply(sapply(f_2_10_tc, function(x) x$errparms), 1, mean, na.rm = TRUE)
+
+  tcf_2_10_error_model_errors <- (tcf_2_10_tc - c(0.5, 0.07)) / c(0.5, 0.07)
+  expect_true(all(abs(tcf_2_10_error_model_errors) < 0.2))
+
+  # When we have 100 replicates in the synthetic data, we can roundtrip
+  # the parameters with < 2% precision
+  f_tc_100_1 <- mkinfit(DFOP, d_100_1[[1]], error_model = "tc", quiet = TRUE)
+  parm_errors_100_1 <- (f_tc_100_1$bparms.optim - parms_DFOP_optim) / parms_DFOP_optim
+  expect_true(all(abs(parm_errors_100_1) < 0.02))
+
+  tcf_100_1_error_model_errors <- (f_tc_100_1$errparms - c(0.5, 0.07)) /
+    c(0.5, 0.07)
+  # We also get a precision of < 2% for the error model components
+  expect_true(all(abs(tcf_100_1_error_model_errors) < 0.02))
+
+  # Parent and two metabolites
+  m_synth_DFOP_lin <- mkinmod(parent = list(type = "DFOP", to = "M1"),
+                             M1 = list(type = "SFO", to = "M2"),
+                             M2 = list(type = "SFO"), use_of_ff = "max",
+                             quiet = TRUE)
+  sampling_times = c(0, 1, 3, 7, 14, 28, 60, 90, 120)
+  parms_DFOP_lin <- c(k1 = 0.2, k2 = 0.02, g = 0.5,
+     f_parent_to_M1 = 0.5, k_M1 = 0.3,
+     f_M1_to_M2 = 0.7, k_M2 = 0.02)
+  d_synth_DFOP_lin <- mkinpredict(m_synth_DFOP_lin,
+     parms_DFOP_lin,
+     c(parent = 100, M1 = 0, M2 = 0),
+     sampling_times)
+  parms_DFOP_lin_optim = c(parent_0 = 100, parms_DFOP_lin)
+
+  d_met_2_15 <- add_err(d_synth_DFOP_lin,
+    sdfunc = function(x) sigma_twocomp(x, 0.5, 0.07),
+    n = 15, reps = 100, digits = 5, LOD = 0.01, seed = 123456)
+
+  # For a single fit, we get a relative error of less than 5% in the error
+  # model components
+  f_met_2_tc_e4 <- mkinfit(m_synth_DFOP_lin, d_met_2_15[[1]], quiet = TRUE,
+    error_model = "tc", error_model_algorithm = "direct")
+  parm_errors_met_2_tc_e4 <- (f_met_2_tc_e4$errparms - c(0.5, 0.07)) / c(0.5, 0.07)
+  expect_true(all(abs(parm_errors_met_2_tc_e4) < 0.05))
+
+  # Doing more takes a lot of computing power
+  skip_on_travis()
+  skip_on_cran()
+  f_met_2_15_tc_e4 <- mmkin(list(m_synth_DFOP_lin), d_met_2_15, quiet = TRUE,
+                            error_model = "tc", cores = n_cores)
+
+  parms_met_2_15_tc_e4 <- apply(sapply(f_met_2_15_tc_e4, function(x) x$bparms.optim), 1, mean)
+  parm_errors_met_2_15_tc_e4 <- (parms_met_2_15_tc_e4[names(parms_DFOP_lin_optim)] -
+                                 parms_DFOP_lin_optim) / parms_DFOP_lin_optim
+  expect_true(all(abs(parm_errors_met_2_15_tc_e4) < 0.015))
+
+  tcf_met_2_15_tc <- apply(sapply(f_met_2_15_tc_e4, function(x) x$errparms), 1, mean, na.rm = TRUE)
+
+  tcf_met_2_15_tc_error_model_errors <- (tcf_met_2_15_tc - c(0.5, 0.07)) /
+    c(0.5, 0.07)
+
+  # Here we get a precision < 10% for retrieving the original error model components
+  # from 15 datasets
+  expect_true(all(abs(tcf_met_2_15_tc_error_model_errors) < 0.10))
+})
+
+test_that("The different error model fitting methods work for parent fits", {
+  skip_on_cran()
+
+  f_9_OLS <- mkinfit("SFO", experimental_data_for_UBA_2019[[9]]$data,
+                     quiet = TRUE)
+  expect_equivalent(round(AIC(f_9_OLS), 2), 137.43)
+
+  f_9_direct <- mkinfit("SFO", experimental_data_for_UBA_2019[[9]]$data,
+    error_model = "tc", error_model_algorithm = "direct", quiet = TRUE)
+  expect_equivalent(round(AIC(f_9_direct), 2), 134.94)
+
+  f_9_twostep <- mkinfit("SFO", experimental_data_for_UBA_2019[[9]]$data,
+    error_model = "tc", error_model_algorithm = "twostep", quiet = TRUE)
+  expect_equivalent(round(AIC(f_9_twostep), 2), 134.94)
+
+  f_9_threestep <- mkinfit("SFO", experimental_data_for_UBA_2019[[9]]$data,
+    error_model = "tc", error_model_algorithm = "threestep", quiet = TRUE)
+  expect_equivalent(round(AIC(f_9_threestep), 2), 139.43)
+
+  f_9_fourstep <- mkinfit("SFO", experimental_data_for_UBA_2019[[9]]$data,
+    error_model = "tc", error_model_algorithm = "fourstep", quiet = TRUE)
+  expect_equivalent(round(AIC(f_9_fourstep), 2), 139.43)
+
+  f_9_IRLS <- mkinfit("SFO", experimental_data_for_UBA_2019[[9]]$data,
+    error_model = "tc", error_model_algorithm = "IRLS", quiet = TRUE)
+  expect_equivalent(round(AIC(f_9_IRLS), 2), 139.43)
+
+  f_9_d_3 <- mkinfit("SFO", experimental_data_for_UBA_2019[[9]]$data,
+    error_model = "tc", error_model_algorithm = "d_3", quiet = TRUE)
+  expect_equivalent(round(AIC(f_9_d_3), 2), 134.94)
+})