1.0.5 patch (#71)

* sparcc cov and tests

* dont use nonexported huge.mb

* some general data checks

* fiddle test params

* removed git comment marker

* remove performance test, update buildignore

.Rbuildignore

@@ -4,3 +4,5 @@
 ^.README\.*
 ^inst/.README\.*
 ^README\.Rmd$
+^inst/extdata\.*
+^inst/TUTORIAL*

DESCRIPTION

@@ -1,6 +1,6 @@
 Package: SpiecEasi
 Title: Sparse Inverse Covariance for Ecological Statistical Inference
-Version: 1.0.2
+Version: 1.0.5
 Authors@R: c(
     person("Zachary", "Kurtz", role = c("aut", "cre"), email="[email protected]"),
     person("Christian", "Mueller", role = "aut"),
@@ -31,4 +31,4 @@ Suggests:
 biocViews:
 License: GPL (>= 2)
 LazyData: true
-RoxygenNote: 6.1.0
+RoxygenNote: 6.1.1

R/SparseICov.R

@@ -54,11 +54,15 @@ sparseiCov <- function(data, method, npn=FALSE, verbose=FALSE, cov.output = TRUE
   if (is.null(args$lambda.min.ratio)) args$lambda.min.ratio <- 1e-3
 
   if (method %in% c("glasso")) {
-    est <- do.call(huge::huge, c(args, list(x=data, method=method, verbose=verbose,
-                                     cov.output = cov.output)))
+    est <- do.call(huge::huge, c(args, list(x=data,
+                                            method=method,
+                                            verbose=verbose,
+                                            cov.output = cov.output)))
 
   } else if (method %in% c('mb')) {
-    est <- do.call(huge::huge.mb, c(args, list(x=data, verbose=verbose)))
+    est <- do.call(huge::huge, c(args, list(x=data,
+                                            method=method,
+                                            verbose=verbose)))
     est$method <- 'mb'
     est$data <- data
     est$sym  <- ifelse(!is.null(args$sym), args$sym, 'or')

R/spaRcc.R

@@ -20,7 +20,6 @@ sparcc <- function(data, iter=20, inner_iter=10, th=.1) {
     covs <- array(unlist(lapply(sparccs, function(x) x$Cov)),
                  c(ncol(data),ncol(data),iter))
     covMed <- apply(covs, 1:2, median)
-
     covMed <- cor2cov(corMed, sqrt(diag(covMed)))
     list(Cov=covMed, Cor=corMed)
 }

R/spiec-easi.R

@@ -19,6 +19,18 @@ spiec.easi <- function(data, ...) {
   return(OTU)
 }
 
+#' @keywords internal
+.data.checks <- function(data) {
+  ## data checks ##
+  if (isTRUE(all.equal(rowSums(data), rep(1L, nrow(data))))) {
+    warning('Data is normalized, but raw counts are expected')
+  }
+
+  if (any(data<0)) {
+    warning('Negative values detected, but raw counts are expected')
+  }
+}
+
 #' @method spiec.easi phyloseq
 #' @rdname spiec.easi
 #' @export
@@ -129,6 +141,9 @@ spiec.easi.default <- function(data, method='glasso', sel.criterion='stars',
                         icov.select.params=pulsar.params, ...) {
 
   args <- list(...)
+
+  .data.checks(data)
+
   if (verbose) msg <- .makeMessage("Applying data transformations...")
   else msg <- .makeMessage('')
 

tests/testthat/test_pulsar.R

@@ -1,14 +1,14 @@
 context('setup')
 
-p <- 20
+p <- 30
 e <- p
 n <- 500
 set.seed(10010)
 g <- make_graph('erdos_renyi', p, e)
 S <- cov2cor(prec2cov(graph2prec(g)))
 X <- exp(rmvnorm(n, rep(0,p), S))
 
-pargs <- list(seed=10010, rep.num=10)
+pargs <- list(seed=10010, rep.num=15)
 
 context("SPIEC-EASI fit")
 lmx  <- .7
@@ -28,9 +28,6 @@ bout.stars <- spiec.easi(X, method='mb', verbose=FALSE, lambda.max=lmx, lambda.m
 bout.bstars <- spiec.easi(X, method='mb', verbose=FALSE, lambda.max=lmx, lambda.min.ratio=lmr, nlambda=nlam, sel.criterion='bstars', pulsar.select='batch', pulsar.params=pargs)
 
 
-# save(out, out.stars, out.bstars, bout.stars, bout.bstars,
-#     file=system.file('testdata', 'sepulsar.RData', package='SpiecEasi'))
-
 test_that("no pulsar has same output", {
   expect_equal(as.matrix(out$est$path[[out.stars$select$stars$opt.index]]),
                as.matrix(out.stars$refit$stars))
@@ -39,8 +36,6 @@ test_that("no pulsar has same output", {
 test_that("stars == bstars", {
   expect_equal(as.matrix(out.bstars$refit$stars),
                as.matrix(out.stars$refit$stars))
-  expect_gt(t1[3], t2[3])
-
 })
 
 test_that("stars == batch stars", {
@@ -58,14 +53,8 @@ test_that("stars == batch bstars", {
                as.matrix(bout.bstars$refit$stars))
 })
 
-
-## TODO: test pulsar params in/out
-
-
 context('spiec.easi getters')
 
-# load(system.file('testdata', 'sepulsar.RData', package='SpiecEasi'))
-
 test_that("Getter API throws errors if no pulsar selection", {
   expect_error(getOptInd(out))
   expect_error(getOptNet(out))

tests/testthat/test_sparcc.R

@@ -7,13 +7,21 @@ set.seed(10010)
 g <- make_graph('erdos_renyi', p, e)
 S <- cov2cor(prec2cov(graph2prec(g)))
 X <- exp(rmvnorm(n, rep(0,p), S))
+X.f <- t(apply(X, 1, norm_to_total))
 
-context("SparCC ouput")
+context("Input data")
+test_that("data is counts", {
+  expect_silent(.data.checks(X))
+  expect_warning(.data.checks(X.f))
+  expect_warning(.data.checks(scale(X)))
+})
+
+context("SparCC")
 test_that("sparcc gives expected output", {
   out <- sparcc(X)
   expect_true(all(c('Cov', 'Cor') %in% names(out)))
   expect_true(all(diag(out$Cor)==1))
-  expect_false(all(out$Cor==out$Cov))
+  expect_false(any(out$Cor==out$Cov))
   expect_equal(out$Cor, cov2cor(out$Cov))
   expect_equal(dim(out$Cor), c(p,p))
   expect_equal(dim(out$Cov), c(p,p))
@@ -31,5 +39,4 @@ test_that('sparccboot gives expected output', {
   expect_lte(max(pvals$pvals, na.rm=TRUE), 1)
   expect_gte(min(pvals$cors,  na.rm=TRUE), -1)
   expect_lte(max(pvals$cors,  na.rm=TRUE), 1)
-
 })