* using log directory ‘/data/localhost/ripley/R/packages/tests-MKL/IsingSampler.Rcheck’
* using R Under development (unstable) (2026-10-28 r90786)
* using platform: x86_64-pc-linux-gnu
* R was compiled by
    gcc (GCC) 16.2.1 20260819 (Red Hat 16.2.1-2)
    GNU Fortran (GCC) 16.2.1 20260819 (Red Hat 16.2.1-2)
* running under: Fedora Linux 44 (Server Edition)
* using session charset: UTF-8
* current time: 2026-10-29 13:53:45 UTC
* using option ‘--no-stop-on-test-error’
* checking for file ‘IsingSampler/DESCRIPTION’ ... OK
* checking extension type ... Package
* this is package ‘IsingSampler’ version ‘0.5.0’
* checking package namespace information ... OK
* checking package dependencies ... OK
* checking if this is a source package ... OK
* checking if there is a namespace ... OK
* checking for executable files ... OK
* checking for hidden files and directories ... OK
* checking for portable file names ... OK
* checking for sufficient/correct file permissions ... OK
* checking whether package ‘IsingSampler’ can be installed ... [16s/31s] OK
* used C++ compiler: ‘g++ (GCC) 16.2.1 20260819 (Red Hat 16.2.1-2)’
* checking package directory ... OK
* checking DESCRIPTION meta-information ... OK
* checking top-level files ... OK
* checking for left-over files ... OK
* checking index information ... OK
* checking package subdirectories ... OK
* checking code files for non-ASCII characters ... OK
* checking R files for syntax errors ... OK
* checking whether the package can be loaded ... OK
* checking whether the package can be loaded with stated dependencies ... OK
* checking whether the package can be unloaded cleanly ... OK
* checking whether the namespace can be loaded with stated dependencies ... OK
* checking whether the namespace can be unloaded cleanly ... OK
* checking loading without being on the library search path ... OK
* checking use of S3 registration ... OK
* checking dependencies in R code ... OK
* checking S3 generic/method consistency ... OK
* checking replacement functions ... OK
* checking foreign function calls ... OK
* checking R code for possible problems ... OK
* checking Rd files ... OK
* checking Rd metadata ... OK
* checking Rd line widths ... OK
* checking Rd cross-references ... OK
* checking for missing documentation entries ... OK
* checking for code/documentation mismatches ... OK
* checking Rd \usage sections ... OK
* checking Rd contents ... OK
* checking for unstated dependencies in examples ... OK
* checking line endings in C/C++/Fortran sources/headers ... OK
* checking pragmas in C/C++ headers and code ... OK
* checking compilation flags used ... OK
* checking compiled code ... OK
* checking examples ... OK
* checking for unstated dependencies in ‘tests’ ... OK
* checking tests ...
  Running ‘test-constrain.R’
  Running ‘test-distribution.R’
  Running ‘test-entropy.R’
  Running ‘test-estimators.R’
  Running ‘test-minsum.R’
  Running ‘test-pseudolikelihood.R’
  Running ‘test-return-types.R’
  Running ‘test-validation.R’
 [14s/18s] ERROR
Running the tests in ‘tests/test-estimators.R’ failed.
Complete output:
  > # Tests for the non-regularized estimators (EstimateIsing*):
  > # (a) the 'beta' argument must actually be used. Convention: the Ising
  > #     measure depends on the parameters only through beta * graph and
  > #     beta * thresholds, so all estimators return the beta = 1 estimates
  > #     divided by beta (for 'pl' this is achieved by optimizing the
  > #     pseudo-likelihood at the supplied beta, which yields the same scaling);
  > # (b) supplying 'responses' explicitly must give the same result as the
  > #     default (missing) call on matching data;
  > # (c) EstimateIsingPL() must warn when a supplied 'responses' coding does
  > #     not match the values present in the data.
  > # Plain stopifnot tests, no package dependencies beyond IsingSampler itself.
  > 
  > library("IsingSampler")
  > 
  > ## Small seeded 3-node test model ---------------------------------------------
  > set.seed(123)
  > N <- 3
  > graph <- matrix(0, N, N)
  > graph[upper.tri(graph)] <- c(0.4, -0.2, 0.6)
  > graph <- graph + t(graph)
  > thresholds <- c(-0.1, 0.1, 0)
  > data <- as.matrix(IsingSampler(500, graph, thresholds, responses = c(0L, 1L)))
  > 
  > ## (a) beta rescaling ----------------------------------------------------------
  > 
  > # Pseudo-likelihood: fitting at beta = 2 optimizes the same objective as a
  > # function of 2 * par, so the estimates must (up to nlm convergence noise)
  > # equal the beta = 1 estimates divided by 2. Generous tolerance for nlm:
  > pl1 <- EstimateIsingPL(data, beta = 1)
  > pl2 <- EstimateIsingPL(data, beta = 2)
  > stopifnot(max(abs(pl2$graph - pl1$graph / 2)) < 1e-3)
  > stopifnot(max(abs(pl2$thresholds - pl1$thresholds / 2)) < 1e-3)
  > 
  > # Univariate logistic regressions: rescaling is exact (deterministic glm):
  > uni1 <- EstimateIsingUni(data, beta = 1)
  > uni2 <- EstimateIsingUni(data, beta = 2)
  > stopifnot(isTRUE(all.equal(uni2$graph, uni1$graph / 2)))
  > stopifnot(isTRUE(all.equal(uni2$thresholds, uni1$thresholds / 2)))
  > 
  > # Loglinear model: rescaling is exact (deterministic loglin/IPF):
  > ll1 <- EstimateIsingLL(data, beta = 1)
  3 iterations: deviation 0.08676137 
  > ll2 <- EstimateIsingLL(data, beta = 2)
  3 iterations: deviation 0.08676137 
  > stopifnot(isTRUE(all.equal(ll2$graph, ll1$graph / 2)))
  > stopifnot(isTRUE(all.equal(ll2$thresholds, ll1$thresholds / 2)))
  > 
  > # Bivariate (multinomial) regressions can be numerically unstable, so only
  > # the beta rescaling arithmetic is exercised here: identical RNG seeds make
  > # the two multinom fits start identically, so the estimates must satisfy the
  > # same divide-by-beta relation (generous tolerance for the optimizer):
  > set.seed(1)
  > bi1 <- EstimateIsingBi(data, beta = 1, trace = FALSE)
  > set.seed(1)
  > bi2 <- EstimateIsingBi(data, beta = 2, trace = FALSE)
  > stopifnot(all(is.finite(bi2$graph)), all(is.finite(bi2$thresholds)))
  > stopifnot(max(abs(bi2$graph - bi1$graph / 2)) < 1e-3)
  > stopifnot(max(abs(bi2$thresholds - bi1$thresholds / 2)) < 1e-3)
  > 
  > # Cross-method sanity check of the convention: beta * estimates should be on
  > # the same (measure) scale regardless of beta, e.g. 2 * pl2 approx 1 * pl1:
  > stopifnot(max(abs(2 * pl2$graph - pl1$graph)) < 2e-3)
  > 
  > ## (b) explicit responses equal the default on matching 0/1 data ---------------
  > plDef <- EstimateIsingPL(data)
  > plExp <- EstimateIsingPL(data, responses = c(0, 1))
  > stopifnot(identical(plDef$graph, plExp$graph))
  > stopifnot(identical(plDef$thresholds, plExp$thresholds))
  > 
  > uniDef <- EstimateIsingUni(data)
  > uniExp <- EstimateIsingUni(data, responses = c(0, 1))
  > stopifnot(identical(uniDef$graph, uniExp$graph))
  Error: identical(uniDef$graph, uniExp$graph) is not TRUE
  Execution halted
Running the tests in ‘tests/test-minsum.R’ failed.
Complete output:
  > # Tests for the minimum sum-score correction of EstimateIsingUni():
  > # 'min_sum' refers to sum scores on the ORIGINAL response coding, but the
  > # row filter used to run after the data were recoded to 0/1, so for data
  > # coded e.g. c(1,2) it filtered the wrong rows (usually none). The filter
  > # now runs on the raw data: for node i a row is biased iff the rest score
  > # is compatible with the truncation when node i takes the high response
  > # but not when it takes the low response. Checks:
  > # (a) on 0/1 data the new raw-scale filter retains exactly the rows the
  > #     old binarized-scale filter (rest != min_sum - 1) retained, per node;
  > # (b) on c(1,2)-coded data (same data recoded, min_sum shifted accordingly)
  > #     the estimator drops the corresponding rows and the estimates agree
  > #     with the 0/1 fit after LinTransform;
  > # (c) the default min_sum = -Inf drops no rows.
  > # Plain stopifnot tests, no package dependencies beyond IsingSampler itself.
  > 
  > library("IsingSampler")
  > 
  > ## Small seeded 4-node test model ---------------------------------------------
  > set.seed(42)
  > N <- 4
  > graph <- matrix(0, N, N)
  > graph[upper.tri(graph)] <- c(0.4, -0.2, 0.6, 0.2, -0.3, 0.5)
  > graph <- graph + t(graph)
  > thresholds <- c(-0.2, 0.1, 0, -0.1)
  > full <- as.matrix(IsingSampler(1000, graph, thresholds, responses = c(0L, 1L)))
  > 
  > # Mimic sum-score truncated data: keep only rows with sum score >= min_sum:
  > min_sum <- 2
  > data01 <- full[rowSums(full) >= min_sum, , drop = FALSE]
  > 
  > # Number of retained rows per fitted node model:
  > nRows <- function(fit) sapply(fit$results, function(m) length(residuals(m)))
  > 
  > ## (a) new raw-scale filter == old filter on 0/1 data --------------------------
  > # On 0/1 data the binarized data equal the raw data, so the old filter
  > # (rest != min_sum - 1 on the binarized data) must be reproduced exactly by
  > # the new condition (rest + 1 >= min_sum & rest + 0 < min_sum on raw data):
  > uni01 <- EstimateIsingUni(data01, min_sum = min_sum)
  > for (i in seq_len(N)){
  +   rest <- rowSums(data01[, -i, drop = FALSE])
  +   keepOld <- which(rest != min_sum - 1)
  +   keepNew <- which(!(rest + 1 >= min_sum & rest + 0 < min_sum))
  +   stopifnot(identical(keepOld, keepNew))
  +   # The estimator must have fitted node i on exactly these rows:
  +   stopifnot(nRows(uni01)[i] == length(keepNew))
  +   # Make the test non-trivial: some rows must actually be dropped:
  +   stopifnot(length(keepNew) < nrow(data01))
  + }
  > 
  > ## (b) c(1,2)-coded data: same rows dropped, estimates agree -------------------
  > # Recode the same data to responses c(1,2); each row's sum score shifts by
  > # the number of nodes, so the equivalent truncation threshold shifts too:
  > data12 <- data01 + 1
  > min_sum12 <- min_sum + N
  > uni12 <- EstimateIsingUni(data12, min_sum = min_sum12)
  > 
  > # Corresponding rows are dropped per node (the old binarized-scale filter
  > # would have dropped none here, since rest scores of 0/1-binarized data
  > # never equal min_sum12 - 1 = 5 with only 3 remaining nodes):
  > stopifnot(identical(nRows(uni12), nRows(uni01)))
  > 
  > # The glm fits run on identical binarized data and identical retained rows,
  > # so mapping the 0/1 estimates to the (1,2) coding via LinTransform must
  > # reproduce the (1,2) fit essentially exactly:
  > trans <- LinTransform(uni01$graph, uni01$thresholds, from = c(0, 1), to = c(1, 2))
  > stopifnot(isTRUE(all.equal(trans$graph, uni12$graph, tolerance = 1e-6)))
  > stopifnot(isTRUE(all.equal(trans$thresholds, uni12$thresholds, tolerance = 1e-6)))
  > 
  > ## (c) default min_sum = -Inf drops no rows ------------------------------------
  > uniDef <- EstimateIsingUni(full)
  > uniInf <- EstimateIsingUni(full, min_sum = -Inf)
  > stopifnot(all(nRows(uniInf) == nrow(full)))
  > stopifnot(identical(uniDef$graph, uniInf$graph))
  Error: identical(uniDef$graph, uniInf$graph) is not TRUE
  Execution halted
* checking PDF version of manual ... OK
* checking HTML version of manual ... OK
* checking for non-standard things in the check directory ... OK
* checking for detritus in the temp directory ... OK
* DONE

Status: 1 ERROR
See
  ‘/data/localhost/ripley/R/packages/tests-MKL/IsingSampler.Rcheck/00check.log’
for details.

Command exited with non-zero status 1
Time 1:52.87, 56.65 + 7.37