Testing and Plotting Procedures for Biostatistics

Contains miscellaneous functions useful in biostatistics, mostly univariate and multivariate testing procedures with a special emphasis on permutation tests. Many functions intend to simplify user's life by shortening existing procedures or by implementing plotting functions that can be used with as many methods from different packages as possible.

News

version 0.9-72 (10-02-2019)

MVA.cv and MVA.cmv improved to return individual prediction probabilities
MVA.cor corrected for PCIA and CIA based on at least one mix analysis
MVA.scores, MVA.cor and MVA.load improved to return more than 2 axes
fourth-root transformation added to multtest.gp

version 0.9-71 (10-01-2019)

least.rect() corrected for >2 levels (printing error)
MVA.cv and MVA.cmv improved to return confusion matrices

version 0.9-70 (05-11-2018)

dunn.test removed (the dunnTest function of the package FSA does the same job)
pairwise.to.groups removed (not very useful)
package dependencies changed to follow migration of mixOmics from CRAN to Bioconductor

version 0.9-69-3 (14-05-2018)

changes in help pages as requested by CRAN

version 0.9-69-2 (29-01-2018)

cochran.qtest corrected for pairwise comparisons with identical groups (thanks to Duncan Hedderley)

version 0.9-69 (03-01-2018)

OR.multinom and test.multinom corrected for two-level responses
change from package 'lsmeans' to 'emmeans': back.lsmeans becomes back.emmeans, rating.lsmeans becomes rating.emmeans
plotresid improved to deal more properly with "quasi" distributions in GLM(M)s

version 0.9-68 (21-08-2017)

miscellaneous small improvements

version 0.9-66 (10-07-2017)

argument thresh of MVA.corplot improved
few linguistic improvements

version 0.9-65 (03-05-2017)

MVA.synt extended to DIABLO
MVA.scores, MVA.cor, MVA.load (hence MVA.plot) extended to DIABLO
new function DIABLO.cv (cross validation for DIABLO)
new function DIABLO.test (significance test for DIABLO)

version 0.9-64 (25-03-2017)

plotresid extended to glmmPQL models
new function pairwise.var.test (pairwise comparisons based on F-tests)
new function pairwise.perm.var.test (pairwise comparisons based on permutational F-tests)

version 0.9-63 (24-02-2017)

back.lsmeans corrected for log transformations, fourth-root transformation added
MVA.plot adapted for GPA
plotresid re-thought

version 0.9-62 (12-12-2016)

se corrected for proportions

version 0.9-61 (31-10-2016)

MVA.cv and MVA.cmv corrected for errors provoked by scaling with nearly constant variables

version 0.9-60 (20-08-2016)

MVA.synt extended to MCA, mix analysis, GPA, 2B-PLS, CCorA, rCCorA, RGCCA, CIA and PCIA
MVA.corplot extended to db-RDA
new function stand (standardization of a data frame based on another data frame)
intergroup variance removed from MVA.synt (not relevant)
new function CDA.cv (cross validation for CDA)
new predict methods for "CDA.cv" objects
new function CDA.test (significance test for CDA)
new function pairwise.CDA.test (pairwise comparisons for CDA)
pairwise.factorfit corrected for CA and CCA computed with the vegan package
new function to.dudi (conversion of some vegan analyses to ade4 format)
MVA.scores (hence MVA.plot) extended to (Generalized) Procrustean superimposition
MVA.cor (hence MVA.plot) extended to Generalized Procrustean superimposition
new function GPA.test (significance test for GPA)
new function cov.test (significance test for covariance)
new function ord.rw (re-computation of an ordination using given row weights)

version 0.9-57 (14-07-2016)

new function pairwise.factorfit (pairwise comparisons of groups displayed on a factorial map)
MVA.anova extended to deal with partial CCA/RDA/db-RDA
MVA.synt extended to db-RDA
new argument 'correct' in chisq.bintest (continuity correction)
new argument 'y' in se (standard error of a proportion)
new function cdf.discrete (Cumulative Distribution Function of a known discrete distribution)
prop.multinom corrected for 1-row matrices
new function wald.ptheo.multinom.test (Wald tests for comparison of proportions to theoretical values)
multinomial.test, multinomial.multcomp and multinomial.theo.multcomp accept factors as response
new argument 'prop' in multinomial.theo.multcomp (results displayed as counts or proportions)
new function dunn.test (Dunn's test)
new function cramer (Cram�r's association coefficient between 2 nominal variables)
new argument 'progress' in all function performing permutation tests (display of the progress bar)
new argument 'scale' in MVA.cv, MVA.cmv and MVA.test (scaling)
new function pairwise.MVA.test (pairwise comparisons based on cross (model) validation)
new function cond.multinom (condition number of the Hessian matrix of a multinomial log-linear model)
new function OR.multinom (odds ratios in multinomial regression)
new function test.multinom (significance test in multinomial regression)

version 0.9-56 (06-06-2016)

MVA.synt extended to COA, RDA, CCA
new argument 'vars' in multtest.gp and plotting method optimized
splitf, MVA.cv, MVA.cmv, MVA.test optimized

version 0.9-55 (23-04-2016)

function multtest renamed multtest.gp
new function multtest.cor (correlation test for multiple variables)
new function loc.slp (slope of a hand-defined line)
new function coord.proj (coordinates of projected points)
functions MVA.scores, MVA.cor and MVA.load are now visible

version 0.9-54 (28-03-2016)

small bugs corrected in MVA.plot
new argument 'adj' in least.rect (correction of confidence intervals and p-values for multiple testing)
new function spearman.cor.multcomp (computation of adjusted confidence interval of Spearman's correlation coefficients for multiple comparisons)
small bug corrected in prop.multinom
new function dummy (creation of dummy responses from a factor)

version 0.9-53 (28-02-2016)

small bugs corrected in MVA.synt
new function multtest (univariate comparison of groups for multiple variables)

version 0.9-52 (15-11-2015)

small bugs corrected in MVA.plot

version 0.9-51 (17-10-2015)

small bugs corrected in MVA.plot
plotresid rewritten, new support for gls models
new arguments xlab and ylab in byf.hist

version 0.9-50 (08-08-2015)

wald.ptheo.test now deals with overdispersion
new function prop.multinom (computation of proportions and SE based on a multinomial log-linear model)
new function wmean (weighted arithmetic mean)
plot1comp.ind, plot1comp.var, s.corcircle2, scat.mix.categorical, scat.mix.numeric and scatter.coa2 removed (the MVA.plot function of this package does the same job and is more generic)
new function splitf (random splitting of data frame, respecting the relative proportions of levels of a factor)
new function MVA.cmv (cross model validation)
new function MVA.cv (cross validation)
DA.confusion and DA.valid removed (the MVA.cmv and MVA.cv functions of this package do the same job and are more generic)
new predict methods for "MVA.cv" and "MVA.cmv" objects
new function MVA.test (significance test based on cross (model) validation)
PLSDA.test removed (the MVA.test function of this package does the same job and is more generic)
PLSDA.ncomp removed (not really useful and the mvr function of the 'pls' package does nearly the same job)
new function MVA.plot (various plots for multivariate analyses)
cor.sparse removed (the MVA.plot function of this package does the same job and is more generic)
LDA.format not exported anymore (the MVA.plot function of this package does the same job and is more generic)
new arguments 'ord' and 'decreasing' in back.lsmeans
new function MVA.anova (wrapper to anova.cca of the 'vegan' package but for type II tests)
perm.t.test corrected for the computation of p-values (underestimated in some particular cases when using a two tailed test)
MVA.synt rewritten

version 0.9-45 (13-03-2015)

histograms with density polygons improved
G.test, G.multcomp and G.theo.multcomp modified to deal with 0 values (but warning that it should not be done)
new functions multinomial.test, multinomial.multcomp and multinomial.theo.multcomp (exact multinomial goodness-of-fit test and pairwise comparisons)
chisq.bintest splitted into chisq.bintest and fisher.bintest (chi-square and Fisher's exact tests for comparing response probabilities)
new function G.bintest (G test for comparing response probabilities)
new functions chisq.theo.bintest, chisq.bin.exp and prop.bin.multcomp (comparison of response probabilities to theoretical values)
new function fp.test (Fligner-Policello test)
new function CvM.test (two-sample Cramer-von Mises test)
new method for perm.t.test
new method for perm.var.test
new functions pcor and pcor.test (equivalents to cor and cor.test for (semi-)partial correlation)
new function pairwise.to.groups (conversion of a pairwise-comparison matrix to letters showing significantly different factor levels)
"mlm" models are now dealt properly by plotresid
pairwise.manova removed (dealt in a much powerful manner by the 'lsmeans' package)
new function pairwise.perm.manova (pairwise comparisons using permutational MANOVAs)
wald.ptheo.test accepts 2-column matrices as response
new function adonis.II (wrapper to adonis of the 'vegan' package but for type II tests)
new predict method for "coadisc" objects (Correspondence Discriminant Analysis performed with discrimin.coa from 'ade4')
DA.valid and DA.confusion can now handle "coadisc" objects (Correspondence Discriminant Analysis performed with discrimin.coa from 'ade4')
new arguments arrows in s.corcircle2
new function scat.cr ("correlation" of variables to axes in MCA or mix analyses)
MVA.synt extended to NSCA
wilcox.signtest now computes the confidence interval of the Hodges-Lehmann estimator (and this estimator)
new functions mood.medtest (Mood's median test) and pairwise.mood.medtest
new function prop.multinom.test (pairwise comparisons of proportions based on multinomial log-linear models)
wilcox.rating.signtest, wilcox.rating.test, kruskal.rating.test, friedman.rating.test, pairwise.wilcox.rating.test and wilcox.paired.rating.multcomp removed (these tests can be problematic with ratings, CLMs (package 'ordinal') are much better)
new function back.lsmeans (back-transformation of LSMeans and SE interval)

version 0.9-41 (15-11-2014)

lsm.basis.clm and lsm.basis.clmm removed (now dealt in a much powerful manner by the 'lsmeans' package)
DA.var removed (the MVA.synt function of this package does the same job and is more generic)
fc.multcomp removed (comparison of slopes dealt in a much powerful manner by the 'lsmeans' package)
plot1comp.ind rewritten to display a more easy-to-interpret graph
byf.normhist removed (not very useful and nearly same thing now done by byf.hist)
MVA.synt extended to MCA and mixed analyses