Nothing
context("Checking wfm")
# expect_true <- isTRUE; expect_equivalent <- all.equal
test_that("wfm out puts a matrix with appropriate attributes",{
wfm1 <- with(DATA, wfm(state, list(sex, adult)))[1:15, ]
wfm2 <- wfm(DATA[["state"]], DATA[["person"]])[1:15, ]
wfm1_c <- structure(c(1, 0, 0, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 1, 1, 0, 0, 0, 0, 1,
1, 0, 1, 1, 2, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0,
0), .Dim = c(15L, 4L), .Dimnames = list(c("about", "already",
"am", "are", "be", "can", "certain", "computer", "distrust",
"do", "dumb", "eat", "fun", "good", "how"), c("f.0", "f.1", "m.0",
"m.1")))
wfm2_c <- structure(c(0, 1, 1, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 0, 1, 1, 1, 1, 0,
0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0, 0, 0, 2, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0), .Dim = c(15L,
5L), .Dimnames = list(c("about", "already", "am", "are", "be",
"can", "certain", "computer", "distrust", "do", "dumb", "eat",
"fun", "good", "how"), c("greg", "researcher", "sally", "sam",
"teacher")))
expect_equivalent(wfm1,wfm1_c)
expect_equivalent(wfm2,wfm2_c)
})
test_that("wfdf out puts a matrix with appropriate attributes",{
wfdf1 <- with(DATA, wfdf(state, list(sex, adult)))[1:15, ]
wfdf2 <- with(DATA, wfdf(state, person))[1:15, ]
wfdf1_c <- structure(list(Words = c("about", "already", "am", "are", "be",
"can", "certain", "computer", "distrust", "do", "dumb", "eat",
"fun", "good", "how"), f.0 = c(1, 0, 0, 1, 1, 1, 1, 0, 0, 0,
0, 0, 0, 0, 1), f.1 = c(0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1, 0), m.0 = c(0, 1, 1, 0, 0, 0, 0, 1, 1, 0, 1, 1, 2, 0, 0),
m.1 = c(0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0)), .Names = c("Words",
"f.0", "f.1", "m.0", "m.1"), row.names = c(NA, 15L), class = c("wfdf",
"t.df", "data.frame"))
wfdf2_c <- structure(list(Words = c("about", "already", "am", "are", "be",
"can", "certain", "computer", "distrust", "do", "dumb", "eat",
"fun", "good", "how"), greg = c(0, 1, 1, 0, 0, 0, 0, 0, 0, 0,
1, 1, 0, 0, 0), researcher = c(0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 1, 0), sally = c(1, 0, 0, 1, 1, 1, 1, 0, 0, 0, 0, 0,
0, 0, 1), sam = c(0, 0, 0, 0, 0, 0, 0, 1, 1, 0, 0, 0, 2, 0, 0
), teacher = c(0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0)), .Names = c("Words",
"greg", "researcher", "sally", "sam", "teacher"), row.names = c(NA,
15L), class = c("wfdf", "t.df", "data.frame"))
expect_equivalent(wfdf1, wfdf1_c)
expect_equivalent(wfdf2, wfdf2_c)
})
test_that("wfm_expanded expands",{
expect_true(2 == sum(rownames(wfm_expanded(wfm(DATA$state,
DATA$person))) == "fun"))
})
test_that("wf_combine combines",{
## wf_combine examples:
#===================
## raw no margins (will work)
x <- wfm(DATA$state, DATA$person)
## raw with margin (will work)
y <- wfdf(DATA$state, DATA$person, margins = TRUE)
## Proportion matrix
z2 <- wfm(DATA$state, DATA$person, output="proportion")
WL1 <- c(y[, 1])
WL2 <- list(c("read", "the", "a"), c("you", "your", "you're"))
WL3 <- list(bob = c("read", "the", "a"), yous = c("you", "your", "you're"))
WL4 <- list(bob = c("read", "the", "a"), yous = c("a", "you", "your", "your're"))
WL5 <- list(yous = c("you", "your", "your're"))
WL6 <- list(c("you", "your", "your're")) #no name so will be called words 1
WL7 <- c("you", "your", "your're")
expect_error(wfm_combine(z2, WL2)) #Won't work not a raw frequency matrix
expect_error(wfm_combine(y, WL4)) #Error
wc1 <- wfm_combine(x, WL2) #Works (raw and no margins)
wc2 <- wfm_combine(y, WL2) #Works (raw with margins)
wc3 <- wfm_combine(y, c("you", "your", "your're"))
wc4 <- wfm_combine(y, WL1)
wc5 <- wfm_combine(y, WL3)
wc6 <- wfm_combine(y, WL5)
wc7 <- wfm_combine(y, WL6)
wc8 <- wfm_combine(y, WL7)
wc1_c <- structure(c(1, 1, 19, 0, 0, 7, 0, 1, 10, 0, 2, 12, 0, 0, 5), .Dim = c(3L,
5L), .Dimnames = list(c("words 1", "words 2", "else.words"),
c("greg", "researcher", "sally", "sam", "teacher")), class = c("wfm",
"true.matrix", "matrix"))
wc2_c <- structure(c(1, 1, 18, 0, 0, 6, 0, 1, 9, 0, 2, 11, 0, 0, 4), .Dim = c(3L,
5L), .Dimnames = list(c("words 1", "words 2", "else.words"),
c("greg", "researcher", "sally", "sam", "teacher")), class = c("wfm",
"true.matrix", "matrix"))
wc3_c <- structure(c(1, 19, 0, 6, 1, 9, 2, 11, 0, 4), .Dim = c(2L, 5L), .Dimnames = list(
c("words", "else.words"), c("greg", "researcher", "sally",
"sam", "teacher")), class = c("wfm", "true.matrix", "matrix"
))
wc4_c <- structure(c(20, 6, 10, 13, 4), .Dim = c(1L, 5L), .Dimnames = list(
"WL1", c("greg", "researcher", "sally", "sam", "teacher")), class = c("wfm",
"true.matrix", "matrix"))
wc5_c <- structure(c(1, 1, 18, 0, 0, 6, 0, 1, 9, 0, 2, 11, 0, 0, 4), .Dim = c(3L,
5L), .Dimnames = list(c("bob", "yous", "else.words"), c("greg",
"researcher", "sally", "sam", "teacher")), class = c("wfm", "true.matrix",
"matrix"))
wc6_c <- structure(c(1, 19, 0, 6, 1, 9, 2, 11, 0, 4), .Dim = c(2L, 5L), .Dimnames = list(
c("yous", "else.words"), c("greg", "researcher", "sally",
"sam", "teacher")), class = c("wfm", "true.matrix", "matrix"
))
wc7_c <- structure(c(1, 19, 0, 6, 1, 9, 2, 11, 0, 4), .Dim = c(2L, 5L), .Dimnames = list(
c("words 1", "else.words"), c("greg", "researcher", "sally",
"sam", "teacher")), class = c("wfm", "true.matrix", "matrix"
))
wc8_c <- structure(c(1, 19, 0, 6, 1, 9, 2, 11, 0, 4), .Dim = c(2L, 5L), .Dimnames = list(
c("WL7", "else.words"), c("greg", "researcher", "sally",
"sam", "teacher")), class = c("wfm", "true.matrix", "matrix"
))
expect_equivalent(wc2, wc2_c)
expect_equivalent(wc3, wc3_c)
expect_equivalent(wc4, wc4_c)
expect_equivalent(wc5, wc5_c)
expect_equivalent(wc6, wc6_c)
expect_equivalent(wc7, wc7_c)
expect_equivalent(wc8, wc8_c)
})
# test_that("weight produces matrices of the right class with right attributes",{
#
# WFM <- with(DATA, wfm(state, list(sex, adult)))
# w_wfm1 <- weight(WFM, "prop")
# w_wfm2 <- weight(WFM, "max")
# w_wfm3 <- weight(WFM, "scaled")
#
# w_wfm1_c <- structure(list(dim = c(43L, 4L), dimnames = list(c("about", "already",
# "am", "are", "be", "can", "certain", "computer", "distrust",
# "do", "dumb", "eat", "f", "fun", "good", "how", "hungry", "i",
# "i'm", "is", "it", "it's", "let's", "liar", "m", "move", "no",
# "not", "on", "shall", "should", "stinks", "talking", "telling",
# "the", "then", "there", "too", "truth", "way", "we", "what",
# "you"), c("f.0", "f.1", "m.0", "m.1")), class = c("weighted_wfm",
# "matrix"), weighting = "prop"), .Names = c("dim", "dimnames",
# "class", "weighting"))
#
# w_wfm2_c <- structure(list(dim = c(43L, 4L), dimnames = list(c("about", "already",
# "am", "are", "be", "can", "certain", "computer", "distrust",
# "do", "dumb", "eat", "f", "fun", "good", "how", "hungry", "i",
# "i'm", "is", "it", "it's", "let's", "liar", "m", "move", "no",
# "not", "on", "shall", "should", "stinks", "talking", "telling",
# "the", "then", "there", "too", "truth", "way", "we", "what",
# "you"), c("f.0", "f.1", "m.0", "m.1")), class = c("weighted_wfm",
# "matrix"), weighting = "max"), .Names = c("dim", "dimnames",
# "class", "weighting"))
#
# w_wfm3_c <- structure(list(dim = c(43L, 4L), dimnames = list(c("about", "already",
# "am", "are", "be", "can", "certain", "computer", "distrust",
# "do", "dumb", "eat", "f", "fun", "good", "how", "hungry", "i",
# "i'm", "is", "it", "it's", "let's", "liar", "m", "move", "no",
# "not", "on", "shall", "should", "stinks", "talking", "telling",
# "the", "then", "there", "too", "truth", "way", "we", "what",
# "you"), c("f.0", "f.1", "m.0", "m.1")), class = c("weighted_wfm",
# "matrix"), weighting = "scaled"), .Names = c("dim", "dimnames",
# "class", "weighting"))
#
# expect_equivalent(attributes(w_wfm1), w_wfm1_c)
# expect_equivalent(attributes(w_wfm2), w_wfm2_c)
# expect_equivalent(attributes(w_wfm3), w_wfm3_c)
#
# })
Any scripts or data that you put into this service are public.
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.