#devtools::test("asremlPlus") context("model_selection") if (Sys.getenv("NOT_CRAN") == "true") require(asreml) library(asremlPlus) cat("#### Test for wheat76 spatial example with asreml42\n") test_that("Wheat_spatial_asreml42", { skip_if_not_installed("asreml") skip_on_cran() library(asreml) library(asremlPlus) library(qqplotr) ## use asremlPlus to analyse the wheat (barley) example from section 8.6 of the asreml manual (Butler et al. 2010) data(Wheat.dat) asreml::asreml.options(extra = 5, ai.sing = TRUE, fail = "soft") #Add row and column covariates tmp.dat <- within(Wheat.dat, { cColumn <- dae::as.numfac(Column) cColumn <- cColumn - mean(unique(cColumn)) cRow <- dae::as.numfac(Row) cRow <- cRow - mean(unique(cRow)) }) current.asr <- do.call(asreml, list(yield ~ Rep + WithinColPairs + Variety, random = ~ Row + Column, residual = ~ Row:Column, data=tmp.dat, maxit = 10)) summary(current.asr)$varcomp info <- infoCriteria(current.asr, IClikelihood = "full") testthat::expect_equal(info$varDF, 3) testthat::expect_lt(abs(info$AIC - 1720.891), 0.10) # Load init fit into an asrtests object current.asrt <- as.asrtests(current.asr, NULL, NULL, IClikelihood = "full", label = "Initial model") testthat::expect_lt(abs(current.asrt$test.summary$AIC - 1720.891), 0.50) # Check for and remove any boundary terms current.asrt <- rmboundary(current.asrt, IClikelihood = "full") #Check term for within Column pairs current.asrt <- changeModelOnIC(current.asrt, dropFixed = "WithinColPairs", label = "Try dropping withinColPairs", IClikelihood = "full") print(current.asrt) #Try corb - worst fit corb.asrt <- addSpatialModelOnIC(current.asrt, spatial.model = "corr", row.covar = "cRow", col.covar = "cColumn", row.factor = "Row", col.factor = "Column", corr.funcs = c("corb", "corb"), corr.orders = c(0,0), IClikelihood = "full") corb.asrt <- rmboundary(corb.asrt, IClikelihood = "full") inf <- infoCriteria(corb.asrt$asreml.obj, IClikelihood = "full") testthat::expect_equal(inf$varDF, 3) testthat::expect_true(abs(inf$AIC - 1718.609 ) < 0.1) #Fit autocorrelation model spatialEach.asrts <- list() spatialEach.asrts[["corr"]] <- addSpatialModelOnIC(current.asrt, spatial.model = "corr", row.covar = "cRow", col.covar = "cColumn", row.factor = "Row", col.factor = "Column", IClikelihood = "full") spatialEach.asrts[["corr"]] <- rmboundary(spatialEach.asrts[["corr"]], IClikelihood = "full") spatialEach.asrts[["TPNCSS"]] <- addSpatialModelOnIC(current.asrt, spatial.model = "TPNCSS", row.covar = "cRow", col.covar = "cColumn", row.factor = "Row", col.factor = "Column", dropRandom = "Row + Column", IClikelihood = "full") spatialEach.asrts[["TPNCSS"]] <- rmboundary(spatialEach.asrts[["TPNCSS"]], IClikelihood = "full") spatialEach.asrts[["TPPSC2"]] <- addSpatialModelOnIC(current.asrt, spatial.model = "TPPS", row.covar = "cRow", col.covar = "cColumn", row.factor = "Row", col.factor = "Column", dropRandom = "Row + Column", degree = c(3,3), difforder = c(2,2), rotateX = TRUE, ngridangles = NULL, asreml.option = "grp", IClikelihood = "full") spatialEach.asrts[["TPPSC2"]] <- rmboundary(spatialEach.asrts[["TPPSC2"]], IClikelihood = "full") spatialEach.asrts[["TPPSL1"]] <- addSpatialModelOnIC(current.asrt, spatial.model = "TPPS", row.covar = "cRow", col.covar = "cColumn", row.factor = "Row", col.factor = "Column", dropRandom = "Row + Column", degree = c(1,1), difforder = c(1,1), asreml.option = "grp", IClikelihood = "full") spatialEach.asrts[["TPPSL1"]] <- rmboundary(spatialEach.asrts[["TPPSL1"]], IClikelihood = "full") infoEach <- do.call(rbind, lapply(spatialEach.asrts, function(asrt) infoCriteria(asrt$asreml.obj, IClikelihood = "full"))) (infoEach) #Choose spatial model spatial.asrts <- chooseSpatialModelOnIC(current.asrt, row.covar = "cRow", col.covar = "cColumn", row.factor = "Row", col.factor = "Column", dropRandom = "Row + Column", rotateX = TRUE, ngridangles = NULL, asreml.option = "mbf", return.asrts = "all") #Note that the fits of addSpatialModelOnIC and chooseSpatialModelOnIC differ for TPPSL1; #The fit for addSpatialModelOnIC has an extra variance parameter, but big changes # on last iteration; Very strange! print(spatial.asrts$spatial.IC) print(spatial.asrts$asrts$TPNCSS) testthat::expect_equal(length(spatial.asrts$asrts), 4) testthat::expect_equal(spatial.asrts$spatial.IC$varDF, c(3,5,6,7,4)) testthat::expect_true(all(abs(spatial.asrts$spatial.IC$AIC - c(1718.609, 1651.317, 1639.489, 1642.838, 1652.157) ) < 1e-02)) testthat::expect_true(all.equal(spatial.asrts$spatial.IC[2:4,], infoEach[1:3 ,-3], tolerance = 0.5)) #theta.opt == c(0,0) because rotation Unswapped testthat::expect_true(all(abs(attr(spatial.asrts$asrts$TPSC2$asreml.obj, which = "theta.opt")[[1]] - c(20.20269, 64.97291)) < 1e-04)) current.asr <- spatial.asrts$asrts$TPNCSS$asreml.obj printFormulae(current.asr) ## Get current fitted asreml object and update to include standardized residuals current.asr <- update(current.asr, aom=TRUE) Wheat.dat$res <- residuals(current.asr, type = "stdCond") Wheat.dat$fit <- fitted(current.asr) ## Do diagnostic checking ### Do residuals-versus-fitted values plot with(Wheat.dat, plot(fit, res)) ### Plot variofaces variofaces(current.asr, V=NULL, units="addtores", maxit=50, update = FALSE, ncores = parallel::detectCores()) ### Plot normal quantile plot ggplot(data = Wheat.dat, mapping = aes(sample = res)) + qqplotr::stat_qq_band(bandType = "ts") + qqplotr::stat_qq_line() + qqplotr::stat_qq_point() + labs(x = "Theoretical Quantiles", y = "Sample Quantiles", title = "Normal probability plot") + theme(plot.title = element_text(size = 12, face = "bold")) + theme_bw() ## Get Variety predictions and all pairwise prediction differences and p-values Var.diffs <- predictPlus(classify = "Variety", asreml.obj=current.asr, error.intervals="halfLeast", wald.tab=current.asrt$wald.tab, sortFactor = "Variety", tables = "predictions") ## Plot the Variety predictions, with halfLSD intervals, and the p-values plotPredictions(Var.diffs$predictions, classify = "Variety", y = "predicted.value", error.intervals = "half") plotPvalues(Var.diffs) })