R/test.R
In vspinu/bnlearn: Bayesian network structure learning, parameter learning and inference
conditional.test = function(x, y, sx, data, test, B, alpha = 1, learning = TRUE) {
# update the test counter when performing structure learning.
if (learning)
increment.test.counter()
sx = sx[sx != ""]
ndata = nrow(data)
df = NULL
if (length(sx) == 0) {
# all unconditional tests need this subsetting, do it here.
datax = minimal.data.frame.column(data, x)
datay = minimal.data.frame.column(data, y)
# Mutual Infomation (chi-square asymptotic distribution)
if (test == "mi") {
par.test = mi.test(datax, datay, ndata, gsquare = TRUE)
statistic = par.test[1]
df = par.test[2]
p.value = pchisq(statistic, df, lower.tail = FALSE)
}#THEN
# Shrinked Mutual Infomation (chi-square asymptotic distribution)
if (test == "mi-sh") {
par.test = shmi.test(datax, datay, gsquare = TRUE)
statistic = par.test[1]
df = par.test[2]
p.value = pchisq(statistic, df, lower.tail = FALSE)
}#THEN
# Pearson's X^2 test (chi-square asymptotic distribution)
else if (test == "x2") {
par.test = x2.test(datax, datay)
statistic = par.test[1]
df = par.test[2]
p.value = pchisq(statistic, df, lower.tail = FALSE)
}#THEN
# Jonckheere-Terpstra Test (asymptotic normal distribution)
else if (test == "jt") {
statistic = jt(datax, datay)
p.value = 2 * pnorm(abs(statistic), lower.tail = FALSE)
}#THEN
# Canonical (Linear) Correlation (Student's t distribution)
else if (test == "cor") {
statistic = fast.cor(datax, datay, ndata)
df = ndata - 2
p.value = pt(abs((statistic * sqrt(ndata - 2) / sqrt(1 - statistic^2))),
df, lower.tail = FALSE) * 2
}#THEN
# Fisher's Z (asymptotic normal distribution)
else if (test == "zf") {
statistic = fast.cor(datax, datay, ndata)
statistic = log((1 + statistic)/(1 - statistic))/2 * sqrt(ndata -3)
p.value = pnorm(abs(statistic), lower.tail = FALSE) * 2
}#THEN
# Mutual Information for Gaussian Data (chi-square asymptotic distribution)
else if (test == "mi-g") {
statistic = mig.test(datax, datay, ndata, gsquare = TRUE)
df = 1
p.value = pchisq(statistic, df, lower.tail = FALSE)
}#THEN
# Shrinked Mutual Information for Gaussian Data (chi-square asymptotic distribution)
else if (test == "mi-g-sh") {
statistic = shmig.test(datax, datay, ndata, gsquare = TRUE)
df = 1
p.value = pchisq(statistic, df, lower.tail = FALSE)
}#THEN
# Mutual Infomation (monte carlo permutation distribution)
else if ((test == "mc-mi") || (test == "smc-mi")) {
perm.test = mc.test(datax, datay, samples = B,
alpha = ifelse(test == "smc-mi", alpha, 1), test = 1L)
statistic = perm.test[1]
p.value = perm.test[2]
}#THEN
# Mutual Infomation (semiparametric)
else if (test == "sp-mi") {
perm.test = mc.test(datax, datay, samples = B,
alpha = ifelse(test == "smc-mi", alpha, 1), test = 6L)
statistic = perm.test[1]
df = perm.test[2]
p.value = pchisq(statistic, df, lower.tail = FALSE)
}#THEN
# Pearson's X^2 test (monte carlo permutation distribution)
else if ((test == "mc-x2") || (test == "smc-x2")) {
perm.test = mc.test(datax, datay, samples = B,
alpha = ifelse(test == "smc-x2", alpha, 1), test = 2L)
statistic = perm.test[1]
p.value = perm.test[2]
}#THEN
# Pearson's X^2 test (semiparametric)
else if (test == "sp-x2") {
perm.test = mc.test(datax, datay, samples = B,
alpha = ifelse(test == "smc-mi", alpha, 1), test = 7L)
statistic = perm.test[1]
df = perm.test[2]
p.value = pchisq(statistic, df, lower.tail = FALSE)
}#THEN
# Mutual Information for Gaussian Data (monte carlo permutation distribution)
else if ((test == "mc-mi-g") || (test == "smc-mi-g")) {
statistic = mig.test(datax, datay, ndata, gsquare = TRUE)
p.value = gmc.test(datax, datay, samples = B,
alpha = ifelse(test == "smc-mi-g", alpha, 1), test = 3L)
}#THEN
# Canonical (Linear) Correlation (monte carlo permutation distribution)
else if ((test == "mc-cor") || (test == "smc-cor")) {
statistic = fast.cor(datax, datay, ndata)
p.value = gmc.test(datax, datay, samples = B,
alpha = ifelse(test == "smc-cor", alpha, 1), test = 4L)
}#THEN
# Fisher's Z (monte carlo permutation distribution)
else if ((test == "mc-zf") || (test == "smc-zf")) {
statistic = fast.cor(datax, datay, ndata)
statistic = log((1 + statistic)/(1 - statistic))/2 * sqrt(ndata -3)
p.value = gmc.test(datax, datay, samples = B,
alpha = ifelse(test == "smc-zf", alpha, 1), test = 5L)
}#THEN
# Jonckheere-Terpstra test (monte carlo permutation distribution)
else if ((test == "mc-jt") || (test == "smc-jt")) {
perm.test = mc.test(datax, datay, samples = B,
alpha = ifelse(test == "smc-jt", alpha, 1), test = 8L)
statistic = perm.test[1]
p.value = perm.test[2]
}#THEN
}#THEN
else {
# build the contingency table only for discrete data.
if (test %in% c(available.discrete.tests, available.ordinal.tests)) {
# if there is only one parent, get it easy.
if (length(sx) == 1)
config = minimal.data.frame.column(data, sx)
else
config = configurations(minimal.data.frame.column(data, sx))
}#THEN
# Conditional Mutual Infomation (chi-square asymptotic distribution)
if (test == "mi") {
datax = minimal.data.frame.column(data, x)
datay = minimal.data.frame.column(data, y)
par.test = cmi.test(datax, datay, config, ndata, gsquare = TRUE)
statistic = par.test[1]
df = par.test[2]
p.value = pchisq(statistic, df, lower.tail = FALSE)
}#THEN
# Shrinked Conditional Mutual Infomation (chi-square asymptotic distribution)
if (test == "mi-sh") {
datax = minimal.data.frame.column(data, x)
datay = minimal.data.frame.column(data, y)
par.test = shcmi.test(datax, datay, config, gsquare = TRUE)
statistic = par.test[1]
df = par.test[2]
p.value = pchisq(statistic, df, lower.tail = FALSE)
}#THEN
# Pearson's X^2 test (chi-square asymptotic distribution)
else if (test == "x2") {
datax = minimal.data.frame.column(data, x)
datay = minimal.data.frame.column(data, y)
par.test = cx2.test(datax, datay, config)
statistic = par.test[1]
df = par.test[2]
p.value = pchisq(statistic, df, lower.tail = FALSE)
}#THEN
# Jonckheere-Terpstra Test (asymptotic normal distribution)
else if (test == "jt") {
datax = minimal.data.frame.column(data, x)
datay = minimal.data.frame.column(data, y)
statistic = cjt(datax, datay, config)
p.value = 2 * pnorm(abs(statistic), lower.tail = FALSE)
}#THEN
# Canonical Partial Correlation (Student's t distribution)
else if (test == "cor") {
# define the degrees of freedom and check them.
df = ndata - 2 - length(sx)
if (df < 1)
stop("trying to do a conditional independence test with zero degrees of freedom.")
statistic = fast.pcor(x, y, sx, data, ndata, strict = !learning)
p.value = pt(abs(statistic * sqrt(df) / sqrt(1 - statistic^2)), df, lower.tail = FALSE) * 2
}#THEN
# Fisher's Z (asymptotic normal distribution)
else if (test == "zf") {
# define the degrees of freedom and check them.
df = ndata - 3 - length(sx)
if (df < 1)
stop("trying to do a conditional independence test with zero degrees of freedom.")
statistic = fast.pcor(x, y, sx, data, ndata, strict = !learning)
statistic = log((1 + statistic)/(1 - statistic))/2 * sqrt(df)
p.value = pnorm(abs(statistic), lower.tail = FALSE) * 2
}#THEN
# Mutual Information for Gaussian Data (chi-square asymptotic distribution)
else if (test == "mi-g") {
statistic = cmig.test(x, y, sx, data, ndata, strict = !learning, gsquare = TRUE)
df = 1
p.value = pchisq(statistic, df, lower.tail = FALSE)
}#THEN
# Shrinked Mutual Information for Gaussian Data (chi-square asymptotic distribution)
else if (test == "mi-g-sh") {
statistic = shcmig.test(x, y, sx, data, ndata, gsquare = TRUE)
df = 1
p.value = pchisq(statistic, df, lower.tail = FALSE)
}#THEN
# Mutual Infomation (monte carlo permutation distribution)
else if ((test == "mc-mi") || (test == "smc-mi")) {
datax = minimal.data.frame.column(data, x)
datay = minimal.data.frame.column(data, y)
perm.test = cmc.test(datax, datay, config, samples = B,
alpha = ifelse(test == "smc-mi", alpha, 1), test = 1L)
statistic = perm.test[1]
p.value = perm.test[2]
}#THEN
# Mutual Infomation (semiparametric)
else if (test == "sp-mi") {
datax = minimal.data.frame.column(data, x)
datay = minimal.data.frame.column(data, y)
perm.test = cmc.test(datax, datay, config, samples = B,
alpha = ifelse(test == "smc-mi", alpha, 1), test = 6L)
statistic = perm.test[1]
df = perm.test[2]
p.value = pchisq(statistic, df, lower.tail = FALSE)
}#THEN
# Pearson's X^2 test (monte carlo permutation distribution)
else if ((test == "mc-x2") || (test == "smc-x2")) {
datax = minimal.data.frame.column(data, x)
datay = minimal.data.frame.column(data, y)
perm.test = cmc.test(datax, datay, config, samples = B,
alpha = ifelse(test == "smc-x2", alpha, 1), test = 2L)
statistic = perm.test[1]
p.value = perm.test[2]
}#THEN
# Pearson's X^2 test (semiparametric)
else if (test == "sp-x2") {
datax = minimal.data.frame.column(data, x)
datay = minimal.data.frame.column(data, y)
perm.test = cmc.test(datax, datay, config, samples = B,
alpha = ifelse(test == "smc-x2", alpha, 1), test = 7L)
statistic = perm.test[1]
df = perm.test[2]
p.value = pchisq(statistic, df, lower.tail = FALSE)
}#THEN
# Mutual Information for Gaussian Data (monte carlo permutation distribution)
else if ((test == "mc-mi-g") || (test == "smc-mi-g")) {
statistic = cmig.test(x, y, sx, data, ndata, strict = !learning, gsquare = TRUE)
p.value = cgmc.test(x, y, sx, data, ndata, samples = B,
alpha = ifelse(test == "smc-mi-g", alpha, 1), test = 3L)
}#THEN
# Canonical Partial Correlation (monte carlo permutation distribution)
else if ((test == "mc-cor") || (test == "smc-cor")) {
statistic = fast.pcor(x, y, sx, data, ndata, strict = !learning)
p.value = cgmc.test(x, y, sx, data, ndata, samples = B,
alpha = ifelse(test == "smc-cor", alpha, 1), test = 4L)
}#THEN
# Fisher's Z (monte carlo permutation distribution)
else if ((test == "mc-zf") || (test == "smc-zf")) {
df = ndata - 3 - length(sx)
statistic = fast.pcor(x, y, sx, data, ndata, strict = !learning)
statistic = log((1 + statistic)/(1 - statistic))/2 * sqrt(df)
p.value = cgmc.test(x, y, sx, data, ndata, samples = B,
alpha = ifelse(test == "smc-zf", alpha, 1), test = 5L)
}#THEN
# Jonckheere-Terpstra test (monte carlo permutation distribution)
else if ((test == "mc-jt") || (test == "smc-jt")) {
datax = minimal.data.frame.column(data, x)
datay = minimal.data.frame.column(data, y)
perm.test = cmc.test(datax, datay, config, samples = B,
alpha = ifelse(test == "smc-jt", alpha, 1), test = 8L)
statistic = perm.test[1]
p.value = perm.test[2]
}#THEN
}#ELSE
if (learning) {
return(p.value)
}#THEN
else {
# build a valid object of class htest.
result = structure(
list(statistic = structure(statistic, names = test),
p.value = p.value,
method = test.labels[test],
null.value = c(value = 0),
alternative = ifelse(test %in% c("cor", "zf", "mc-cor", "mc-zf"),
"two.sided", "greater"),
data.name = paste(x, "~", y,
ifelse(length(sx) > 0, "|", ""),
paste(sx, collapse = " + "))
), class = "htest")
# degrees of freedom.
if (!is.null(df))
result$parameter = structure(df, names = "df")
# number of permutations.
if (!is.null(B))
result$parameter = c(result$parameter, structure(B, names = "Monte Carlo samples"))
return(result)
}#ELSE
}#CONDITIONAL.TEST
# Mutual Information (discrete data)
mi.test = function(x, y, ndata, gsquare = TRUE) {
s = .Call("mi",
x = x,
y = y,
gsquare = gsquare)
}#MI.TEST
# Shrinked Mutual Information (discrete data)
shmi.test = function(x, y, gsquare = TRUE) {
s = .Call("shmi",
x = x,
y = y,
gsquare = gsquare)
}#SHMI.TEST
# Monte Carlo (discrete data)
mc.test = function(x, y, samples, alpha, test) {
.Call("mcarlo",
x = x,
y = y,
samples = samples,
test = test,
alpha = alpha)
}#MC.TEST
# Conditional Mutual Information (discrete data)
cmi.test = function(x, y, z, ndata, gsquare = TRUE) {
s = .Call("cmi",
x = x,
y = y,
z = z,
gsquare = gsquare)
}#CMI.TEST
# Shrinked Conditional Mutual Information (discrete data)
shcmi.test = function(x, y, z, gsquare = TRUE) {
s = .Call("shcmi",
x = x,
y = y,
z = z,
gsquare = gsquare)
}#SHCMI.TEST
# Conditional Monte Carlo (discrete data)
cmc.test = function(x, y, z, samples, alpha, test) {
.Call("cmcarlo",
x = x,
y = y,
z = z,
samples = samples,
test = test,
alpha = alpha)
}#CMC.TEST
# Mutual Information (gaussian data)
mig.test = function(x, y, ndata, gsquare = TRUE) {
s = .Call("mig",
x = x,
y = y,
length = ndata)
ifelse(gsquare, 2 * ndata * s, s)
}#MIG.TEST
# Shrinked Mutual Information (gaussian data)
shmig.test = function(x, y, ndata, gsquare = TRUE) {
s = - 0.5 * log(1 - fast.shcor(x, y, ndata)^2)
ifelse(gsquare, 2 * ndata * s, s)
}#SHMIG.TEST
# Conditional Mutual Information (gaussian data)
cmig.test = function(x, y, z, data, ndata, strict, gsquare = TRUE) {
s = - 0.5 * log(1 - fast.pcor(x, y, z, data, ndata, strict)^2)
ifelse(gsquare, 2 * ndata * s, s)
}#CMIG.TEST
# Shrinked Conditional Mutual Information (gaussian data)
shcmig.test = function(x, y, z, data, ndata, gsquare = TRUE) {
s = - 0.5 * log(1 - fast.shpcor(x, y, z, data, ndata, strict = TRUE)^2)
ifelse(gsquare, 2 * ndata * s, s)
}#SHCMIG.TEST
# Monte Carlo (gaussian data)
gmc.test = function(x, y, samples, alpha, test) {
.Call("gauss_mcarlo",
x = x,
y = y,
samples = samples,
test = test,
alpha = alpha)
}#GMC.TEST
# Conditional Monte Carlo (gaussian data)
cgmc.test = function(x, y, sx, data, ndata, samples, alpha, test) {
.Call("gauss_cmcarlo",
data = minimal.data.frame.column(data, c(x, y, sx)),
length = ndata,
samples = samples,
test = test,
alpha = alpha)
}#CGMC.TEST
# Pearson's X^2 test (discrete data)
x2.test = function(x, y) {
.Call("x2",
x = x,
y = y)
}#X2.TEST
# Pearson's Conditional X^2 test (discrete data)
cx2.test = function(x, y, z) {
.Call("cx2",
x = x,
y = y,
z = z)
}#CX2.TEST
# Fast implementation of the linear correlation coefficient.
fast.cor = function(x, y, ndata) {
.Call("fast_cor",
x = x,
y = y,
length = ndata)
}#FAST.COR
# Fast implementation of the partial correlation coefficient.
fast.pcor = function(x, y, sx, data, ndata, strict) {
.Call("fast_pcor",
data = minimal.data.frame.column(data, c(x, y, sx)),
length = ndata,
shrinkage = FALSE,
strict = strict)
}#FAST.PCOR
# Fast implementation of the shrinked linear correlation coefficient.
fast.shcor = function(x, y, ndata) {
.Call("fast_shcor",
x = x,
y = y,
length = ndata)
}#FAST.SHCOR
# Fast implementation of the shrinked partial correlation coefficient.
fast.shpcor = function(x, y, sx, data, ndata, strict) {
.Call("fast_pcor",
data = minimal.data.frame.column(data, c(x, y, sx)),
length = ndata,
shrinkage = TRUE,
strict = strict)
}#FAST.SHPCOR
# Jonckheere-Terpstra test.
jt = function(x, y) {
.Call("jt",
x = x,
y = y)
}#JT
# Conditional Jonckheere-Terpstra test.
cjt = function(x, y, z) {
.Call("cjt",
x = x,
y = y,
z = z)
}#CJT
vspinu/bnlearn documentation built on May 3, 2019, 7:08 p.m.
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
conditional.test = function(x, y, sx, data, test, B, alpha = 1, learning = TRUE) {
# update the test counter when performing structure learning.
if (learning)
increment.test.counter()
sx = sx[sx != ""]
ndata = nrow(data)
df = NULL
if (length(sx) == 0) {
# all unconditional tests need this subsetting, do it here.
datax = minimal.data.frame.column(data, x)
datay = minimal.data.frame.column(data, y)
# Mutual Infomation (chi-square asymptotic distribution)
if (test == "mi") {
par.test = mi.test(datax, datay, ndata, gsquare = TRUE)
statistic = par.test[1]
df = par.test[2]
p.value = pchisq(statistic, df, lower.tail = FALSE)
}#THEN
# Shrinked Mutual Infomation (chi-square asymptotic distribution)
if (test == "mi-sh") {
par.test = shmi.test(datax, datay, gsquare = TRUE)
statistic = par.test[1]
df = par.test[2]
p.value = pchisq(statistic, df, lower.tail = FALSE)
}#THEN
# Pearson's X^2 test (chi-square asymptotic distribution)
else if (test == "x2") {
par.test = x2.test(datax, datay)
statistic = par.test[1]
df = par.test[2]
p.value = pchisq(statistic, df, lower.tail = FALSE)
}#THEN
# Jonckheere-Terpstra Test (asymptotic normal distribution)
else if (test == "jt") {
statistic = jt(datax, datay)
p.value = 2 * pnorm(abs(statistic), lower.tail = FALSE)
}#THEN
# Canonical (Linear) Correlation (Student's t distribution)
else if (test == "cor") {
statistic = fast.cor(datax, datay, ndata)
df = ndata - 2
p.value = pt(abs((statistic * sqrt(ndata - 2) / sqrt(1 - statistic^2))),
df, lower.tail = FALSE) * 2
}#THEN
# Fisher's Z (asymptotic normal distribution)
else if (test == "zf") {
statistic = fast.cor(datax, datay, ndata)
statistic = log((1 + statistic)/(1 - statistic))/2 * sqrt(ndata -3)
p.value = pnorm(abs(statistic), lower.tail = FALSE) * 2
}#THEN
# Mutual Information for Gaussian Data (chi-square asymptotic distribution)
else if (test == "mi-g") {
statistic = mig.test(datax, datay, ndata, gsquare = TRUE)
df = 1
p.value = pchisq(statistic, df, lower.tail = FALSE)
}#THEN
# Shrinked Mutual Information for Gaussian Data (chi-square asymptotic distribution)
else if (test == "mi-g-sh") {
statistic = shmig.test(datax, datay, ndata, gsquare = TRUE)
df = 1
p.value = pchisq(statistic, df, lower.tail = FALSE)
}#THEN
# Mutual Infomation (monte carlo permutation distribution)
else if ((test == "mc-mi") || (test == "smc-mi")) {
perm.test = mc.test(datax, datay, samples = B,
alpha = ifelse(test == "smc-mi", alpha, 1), test = 1L)
statistic = perm.test[1]
p.value = perm.test[2]
}#THEN
# Mutual Infomation (semiparametric)
else if (test == "sp-mi") {
perm.test = mc.test(datax, datay, samples = B,
alpha = ifelse(test == "smc-mi", alpha, 1), test = 6L)
statistic = perm.test[1]
df = perm.test[2]
p.value = pchisq(statistic, df, lower.tail = FALSE)
}#THEN
# Pearson's X^2 test (monte carlo permutation distribution)
else if ((test == "mc-x2") || (test == "smc-x2")) {
perm.test = mc.test(datax, datay, samples = B,
alpha = ifelse(test == "smc-x2", alpha, 1), test = 2L)
statistic = perm.test[1]
p.value = perm.test[2]
}#THEN
# Pearson's X^2 test (semiparametric)
else if (test == "sp-x2") {
perm.test = mc.test(datax, datay, samples = B,
alpha = ifelse(test == "smc-mi", alpha, 1), test = 7L)
statistic = perm.test[1]
df = perm.test[2]
p.value = pchisq(statistic, df, lower.tail = FALSE)
}#THEN
# Mutual Information for Gaussian Data (monte carlo permutation distribution)
else if ((test == "mc-mi-g") || (test == "smc-mi-g")) {
statistic = mig.test(datax, datay, ndata, gsquare = TRUE)
p.value = gmc.test(datax, datay, samples = B,
alpha = ifelse(test == "smc-mi-g", alpha, 1), test = 3L)
}#THEN
# Canonical (Linear) Correlation (monte carlo permutation distribution)
else if ((test == "mc-cor") || (test == "smc-cor")) {
statistic = fast.cor(datax, datay, ndata)
p.value = gmc.test(datax, datay, samples = B,
alpha = ifelse(test == "smc-cor", alpha, 1), test = 4L)
}#THEN
# Fisher's Z (monte carlo permutation distribution)
else if ((test == "mc-zf") || (test == "smc-zf")) {
statistic = fast.cor(datax, datay, ndata)
statistic = log((1 + statistic)/(1 - statistic))/2 * sqrt(ndata -3)
p.value = gmc.test(datax, datay, samples = B,
alpha = ifelse(test == "smc-zf", alpha, 1), test = 5L)
}#THEN
# Jonckheere-Terpstra test (monte carlo permutation distribution)
else if ((test == "mc-jt") || (test == "smc-jt")) {
perm.test = mc.test(datax, datay, samples = B,
alpha = ifelse(test == "smc-jt", alpha, 1), test = 8L)
statistic = perm.test[1]
p.value = perm.test[2]
}#THEN
}#THEN
else {
# build the contingency table only for discrete data.
if (test %in% c(available.discrete.tests, available.ordinal.tests)) {
# if there is only one parent, get it easy.
if (length(sx) == 1)
config = minimal.data.frame.column(data, sx)
else
config = configurations(minimal.data.frame.column(data, sx))
}#THEN
# Conditional Mutual Infomation (chi-square asymptotic distribution)
if (test == "mi") {
datax = minimal.data.frame.column(data, x)
datay = minimal.data.frame.column(data, y)
par.test = cmi.test(datax, datay, config, ndata, gsquare = TRUE)
statistic = par.test[1]
df = par.test[2]
p.value = pchisq(statistic, df, lower.tail = FALSE)
}#THEN
# Shrinked Conditional Mutual Infomation (chi-square asymptotic distribution)
if (test == "mi-sh") {
datax = minimal.data.frame.column(data, x)
datay = minimal.data.frame.column(data, y)
par.test = shcmi.test(datax, datay, config, gsquare = TRUE)
statistic = par.test[1]
df = par.test[2]
p.value = pchisq(statistic, df, lower.tail = FALSE)
}#THEN
# Pearson's X^2 test (chi-square asymptotic distribution)
else if (test == "x2") {
datax = minimal.data.frame.column(data, x)
datay = minimal.data.frame.column(data, y)
par.test = cx2.test(datax, datay, config)
statistic = par.test[1]
df = par.test[2]
p.value = pchisq(statistic, df, lower.tail = FALSE)
}#THEN
# Jonckheere-Terpstra Test (asymptotic normal distribution)
else if (test == "jt") {
datax = minimal.data.frame.column(data, x)
datay = minimal.data.frame.column(data, y)
statistic = cjt(datax, datay, config)
p.value = 2 * pnorm(abs(statistic), lower.tail = FALSE)
}#THEN
# Canonical Partial Correlation (Student's t distribution)
else if (test == "cor") {
# define the degrees of freedom and check them.
df = ndata - 2 - length(sx)
if (df < 1)
stop("trying to do a conditional independence test with zero degrees of freedom.")
statistic = fast.pcor(x, y, sx, data, ndata, strict = !learning)
p.value = pt(abs(statistic * sqrt(df) / sqrt(1 - statistic^2)), df, lower.tail = FALSE) * 2
}#THEN
# Fisher's Z (asymptotic normal distribution)
else if (test == "zf") {
# define the degrees of freedom and check them.
df = ndata - 3 - length(sx)
if (df < 1)
stop("trying to do a conditional independence test with zero degrees of freedom.")
statistic = fast.pcor(x, y, sx, data, ndata, strict = !learning)
statistic = log((1 + statistic)/(1 - statistic))/2 * sqrt(df)
p.value = pnorm(abs(statistic), lower.tail = FALSE) * 2
}#THEN
# Mutual Information for Gaussian Data (chi-square asymptotic distribution)
else if (test == "mi-g") {
statistic = cmig.test(x, y, sx, data, ndata, strict = !learning, gsquare = TRUE)
df = 1
p.value = pchisq(statistic, df, lower.tail = FALSE)
}#THEN
# Shrinked Mutual Information for Gaussian Data (chi-square asymptotic distribution)
else if (test == "mi-g-sh") {
statistic = shcmig.test(x, y, sx, data, ndata, gsquare = TRUE)
df = 1
p.value = pchisq(statistic, df, lower.tail = FALSE)
}#THEN
# Mutual Infomation (monte carlo permutation distribution)
else if ((test == "mc-mi") || (test == "smc-mi")) {
datax = minimal.data.frame.column(data, x)
datay = minimal.data.frame.column(data, y)
perm.test = cmc.test(datax, datay, config, samples = B,
alpha = ifelse(test == "smc-mi", alpha, 1), test = 1L)
statistic = perm.test[1]
p.value = perm.test[2]
}#THEN
# Mutual Infomation (semiparametric)
else if (test == "sp-mi") {
datax = minimal.data.frame.column(data, x)
datay = minimal.data.frame.column(data, y)
perm.test = cmc.test(datax, datay, config, samples = B,
alpha = ifelse(test == "smc-mi", alpha, 1), test = 6L)
statistic = perm.test[1]
df = perm.test[2]
p.value = pchisq(statistic, df, lower.tail = FALSE)
}#THEN
# Pearson's X^2 test (monte carlo permutation distribution)
else if ((test == "mc-x2") || (test == "smc-x2")) {
datax = minimal.data.frame.column(data, x)
datay = minimal.data.frame.column(data, y)
perm.test = cmc.test(datax, datay, config, samples = B,
alpha = ifelse(test == "smc-x2", alpha, 1), test = 2L)
statistic = perm.test[1]
p.value = perm.test[2]
}#THEN
# Pearson's X^2 test (semiparametric)
else if (test == "sp-x2") {
datax = minimal.data.frame.column(data, x)
datay = minimal.data.frame.column(data, y)
perm.test = cmc.test(datax, datay, config, samples = B,
alpha = ifelse(test == "smc-x2", alpha, 1), test = 7L)
statistic = perm.test[1]
df = perm.test[2]
p.value = pchisq(statistic, df, lower.tail = FALSE)
}#THEN
# Mutual Information for Gaussian Data (monte carlo permutation distribution)
else if ((test == "mc-mi-g") || (test == "smc-mi-g")) {
statistic = cmig.test(x, y, sx, data, ndata, strict = !learning, gsquare = TRUE)
p.value = cgmc.test(x, y, sx, data, ndata, samples = B,
alpha = ifelse(test == "smc-mi-g", alpha, 1), test = 3L)
}#THEN
# Canonical Partial Correlation (monte carlo permutation distribution)
else if ((test == "mc-cor") || (test == "smc-cor")) {
statistic = fast.pcor(x, y, sx, data, ndata, strict = !learning)
p.value = cgmc.test(x, y, sx, data, ndata, samples = B,
alpha = ifelse(test == "smc-cor", alpha, 1), test = 4L)
}#THEN
# Fisher's Z (monte carlo permutation distribution)
else if ((test == "mc-zf") || (test == "smc-zf")) {
df = ndata - 3 - length(sx)
statistic = fast.pcor(x, y, sx, data, ndata, strict = !learning)
statistic = log((1 + statistic)/(1 - statistic))/2 * sqrt(df)
p.value = cgmc.test(x, y, sx, data, ndata, samples = B,
alpha = ifelse(test == "smc-zf", alpha, 1), test = 5L)
}#THEN
# Jonckheere-Terpstra test (monte carlo permutation distribution)
else if ((test == "mc-jt") || (test == "smc-jt")) {
datax = minimal.data.frame.column(data, x)
datay = minimal.data.frame.column(data, y)
perm.test = cmc.test(datax, datay, config, samples = B,
alpha = ifelse(test == "smc-jt", alpha, 1), test = 8L)
statistic = perm.test[1]
p.value = perm.test[2]
}#THEN
}#ELSE
if (learning) {
return(p.value)
}#THEN
else {
# build a valid object of class htest.
result = structure(
list(statistic = structure(statistic, names = test),
p.value = p.value,
method = test.labels[test],
null.value = c(value = 0),
alternative = ifelse(test %in% c("cor", "zf", "mc-cor", "mc-zf"),
"two.sided", "greater"),
data.name = paste(x, "~", y,
ifelse(length(sx) > 0, "|", ""),
paste(sx, collapse = " + "))
), class = "htest")
# degrees of freedom.
if (!is.null(df))
result$parameter = structure(df, names = "df")
# number of permutations.
if (!is.null(B))
result$parameter = c(result$parameter, structure(B, names = "Monte Carlo samples"))
return(result)
}#ELSE
}#CONDITIONAL.TEST
# Mutual Information (discrete data)
mi.test = function(x, y, ndata, gsquare = TRUE) {
s = .Call("mi",
x = x,
y = y,
gsquare = gsquare)
}#MI.TEST
# Shrinked Mutual Information (discrete data)
shmi.test = function(x, y, gsquare = TRUE) {
s = .Call("shmi",
x = x,
y = y,
gsquare = gsquare)
}#SHMI.TEST
# Monte Carlo (discrete data)
mc.test = function(x, y, samples, alpha, test) {
.Call("mcarlo",
x = x,
y = y,
samples = samples,
test = test,
alpha = alpha)
}#MC.TEST
# Conditional Mutual Information (discrete data)
cmi.test = function(x, y, z, ndata, gsquare = TRUE) {
s = .Call("cmi",
x = x,
y = y,
z = z,
gsquare = gsquare)
}#CMI.TEST
# Shrinked Conditional Mutual Information (discrete data)
shcmi.test = function(x, y, z, gsquare = TRUE) {
s = .Call("shcmi",
x = x,
y = y,
z = z,
gsquare = gsquare)
}#SHCMI.TEST
# Conditional Monte Carlo (discrete data)
cmc.test = function(x, y, z, samples, alpha, test) {
.Call("cmcarlo",
x = x,
y = y,
z = z,
samples = samples,
test = test,
alpha = alpha)
}#CMC.TEST
# Mutual Information (gaussian data)
mig.test = function(x, y, ndata, gsquare = TRUE) {
s = .Call("mig",
x = x,
y = y,
length = ndata)
ifelse(gsquare, 2 * ndata * s, s)
}#MIG.TEST
# Shrinked Mutual Information (gaussian data)
shmig.test = function(x, y, ndata, gsquare = TRUE) {
s = - 0.5 * log(1 - fast.shcor(x, y, ndata)^2)
ifelse(gsquare, 2 * ndata * s, s)
}#SHMIG.TEST
# Conditional Mutual Information (gaussian data)
cmig.test = function(x, y, z, data, ndata, strict, gsquare = TRUE) {
s = - 0.5 * log(1 - fast.pcor(x, y, z, data, ndata, strict)^2)
ifelse(gsquare, 2 * ndata * s, s)
}#CMIG.TEST
# Shrinked Conditional Mutual Information (gaussian data)
shcmig.test = function(x, y, z, data, ndata, gsquare = TRUE) {
s = - 0.5 * log(1 - fast.shpcor(x, y, z, data, ndata, strict = TRUE)^2)
ifelse(gsquare, 2 * ndata * s, s)
}#SHCMIG.TEST
# Monte Carlo (gaussian data)
gmc.test = function(x, y, samples, alpha, test) {
.Call("gauss_mcarlo",
x = x,
y = y,
samples = samples,
test = test,
alpha = alpha)
}#GMC.TEST
# Conditional Monte Carlo (gaussian data)
cgmc.test = function(x, y, sx, data, ndata, samples, alpha, test) {
.Call("gauss_cmcarlo",
data = minimal.data.frame.column(data, c(x, y, sx)),
length = ndata,
samples = samples,
test = test,
alpha = alpha)
}#CGMC.TEST
# Pearson's X^2 test (discrete data)
x2.test = function(x, y) {
.Call("x2",
x = x,
y = y)
}#X2.TEST
# Pearson's Conditional X^2 test (discrete data)
cx2.test = function(x, y, z) {
.Call("cx2",
x = x,
y = y,
z = z)
}#CX2.TEST
# Fast implementation of the linear correlation coefficient.
fast.cor = function(x, y, ndata) {
.Call("fast_cor",
x = x,
y = y,
length = ndata)
}#FAST.COR
# Fast implementation of the partial correlation coefficient.
fast.pcor = function(x, y, sx, data, ndata, strict) {
.Call("fast_pcor",
data = minimal.data.frame.column(data, c(x, y, sx)),
length = ndata,
shrinkage = FALSE,
strict = strict)
}#FAST.PCOR
# Fast implementation of the shrinked linear correlation coefficient.
fast.shcor = function(x, y, ndata) {
.Call("fast_shcor",
x = x,
y = y,
length = ndata)
}#FAST.SHCOR
# Fast implementation of the shrinked partial correlation coefficient.
fast.shpcor = function(x, y, sx, data, ndata, strict) {
.Call("fast_pcor",
data = minimal.data.frame.column(data, c(x, y, sx)),
length = ndata,
shrinkage = TRUE,
strict = strict)
}#FAST.SHPCOR
# Jonckheere-Terpstra test.
jt = function(x, y) {
.Call("jt",
x = x,
y = y)
}#JT
# Conditional Jonckheere-Terpstra test.
cjt = function(x, y, z) {
.Call("cjt",
x = x,
y = y,
z = z)
}#CJT
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
Embedding an R snippet on your website
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.