tests/testthat/test-GroupICA.R

In iTensor: ICA-Based Matrix/Tensor Decomposition

# Simulation Datasets
d <- 6
X1 <- matrix(runif(200 * d), nrow = 200, ncol = d)
X2 <- matrix(runif(150 * d), nrow = 150, ncol = d)
X3 <- matrix(runif(250 * d), nrow = 250, ncol = d)

Xs <- list(X1 = X1, X2 = X2, X3 = X3)

## Perform GroupICA against Simulation Datasets
J1 <- d
out.pooled <- GroupICA(Xs, J1=J1, algorithm="pooled", ica.algorithm = "FastICA")
# out.Calhoun2009 <- GroupICA(Xs, J1=J1, algorithm="Calhoun2009", ica.algorithm = "FastICA")
# out.Pfister2018 <- GroupICA(Xs, J1=J1, algorithm="Pfister2018", ica.algorithm = "FastICA")

## Test Output object / type
### Test O-1: Object
expect_identical(is.list(out.pooled), TRUE)
# expect_identical(is.list(out.Calhoun2009), TRUE)
# expect_identical(is.list(out.Pfister2018), TRUE)

### Test O-2: Object Names
expect_identical(names(out.pooled),
    c("A", "Ss", "RecError", "RelChange"))
# expect_identical(names(out.Calhoun2009),
#     c("A", "Ss", "RecError", "RelChange"))
# expect_identical(names(out.Pfister2018),
#     c("A", "Ss", "RecError", "RelChange"))

### Test O-3: A
expect_equal(
    dim(out.pooled$A),
    c(ncol(Xs[[1]]), J1))
# expect_equal(
#     dim(out.Calhoun2009$A),
#     c(ncol(Xs[[1]]), J1))
# expect_equal(
#     dim(out.Pfister2018$A),
#     c(ncol(Xs[[1]]), J1))

expect_true(is.matrix(out.pooled$A))
# expect_true(is.matrix(out.Calhoun2009$A))
# expect_true(is.matrix(out.Pfister2018$A))

expect_equal(dim(out.pooled$A), c(ncol(Xs[[1]]), J1))
# expect_equal(dim(out.Calhoun2009$A), c(ncol(Xs[[1]]), J1))
# expect_equal(dim(out.Pfister2018$A), c(ncol(Xs[[1]]), J1))

### Test O-4: Ss
expect_identical(is.matrix(out.pooled$Ss[[1]]), TRUE)
expect_identical(is.matrix(out.pooled$Ss[[2]]), TRUE)
# expect_identical(is.matrix(out.Calhoun2009$Ss[[1]]), TRUE)
# expect_identical(is.matrix(out.Calhoun2009$Ss[[2]]), TRUE)
# expect_identical(is.matrix(out.Pfister2018$Ss[[1]]), TRUE)
# expect_identical(is.matrix(out.Pfister2018$Ss[[2]]), TRUE)

expect_equal(dim(out.pooled$Ss[[1]]), c(nrow(Xs[[1]]), J1))
expect_equal(dim(out.pooled$Ss[[2]]), c(nrow(Xs[[2]]), J1))
# expect_equal(dim(out.Calhoun2009$Ss[[1]]), c(nrow(Xs[[1]]), J1))
# expect_equal(dim(out.Calhoun2009$Ss[[2]]), c(nrow(Xs[[2]]), J1))
# expect_equal(dim(out.Pfister2018$Ss[[1]]), c(nrow(Xs[[1]]), J1))
# expect_equal(dim(out.Pfister2018$Ss[[2]]), c(nrow(Xs[[2]]), J1))

### Test O-5: RecError
expect_identical(is.null(out.pooled$RecError), TRUE)
# expect_identical(is.null(out.Calhoun2009$RecError), TRUE)
# expect_identical(is.null(out.Pfister2018$RecError), TRUE)

# JADE does not return relative change
expect_identical(is.null(out.pooled$RelChange), TRUE)
# expect_identical(is.null(out.Calhoun2009$RelChange), TRUE)
# expect_identical(is.null(out.Pfister2018$RelChange), TRUE)

## Test Error
### Test E-1: X
expect_error(GroupICA(X, J1=J1))

### Test E-2: J1
expect_error(GroupICA(Xs, J1="5"))
expect_error(GroupICA(Xs, J1=c(2,4)))
expect_error(GroupICA(Xs, J1=10^10))

### Test E-3: J2
expect_error(GroupICA(Xs, J2="5"))
expect_error(GroupICA(Xs, J2=c(2,4)))
expect_error(GroupICA(Xs, J2=10^10))

### Test E-4: algorithm
expect_error(GroupICA(Xs, algorithm="poooled"))

### Test E-5: ica.algorithm
expect_error(GroupICA(Xs, ica.algorithm="JAAE"))

### Test E-6: num.iter
expect_error(GroupICA(Xs, J=J, num_iter="100"))
expect_error(GroupICA(Xs, J=J, num_iter=-1))

### Test E-7: thr
expect_error(GroupICA(X, J=J, thr="0.1"))

### Test E-8: verbose
expect_error(GroupICA(X, J=J, verbose="verbose"))