Skip to content

Commit

Permalink
update testing
Browse files Browse the repository at this point in the history 
no errors now in devtools::check()
  • Loading branch information
@qinyun-lin
qinyun-lin committed Mar 14, 2024
1 parent fba85b9 commit 04fbc43
Show file tree
Hide file tree
Showing 4 changed files with 249 additions and 70 deletions.
63 changes: 39 additions & 24 deletions tests/testthat/test-konfound.R
View file
Original file line number Diff line number Diff line change
@@ -1,34 +1,49 @@
context("Checking konfound")

library(forcats)
library(lme4)

m1 <- lm(mpg ~ wt + hp, data = mtcars)
output1 <- konfound(m1, wt, to_return = "raw_output")

# d <- forcats::gss_cat
# d$married <- ifelse(d$marital == "Married", 1, 0)
# m2 <- glm(married ~ age, data = d, family = binomial(link = "logit"))
# output2 <- konfound(m2, age, to_return = "raw_output")

m3 <- fm1 <- lme4::lmer(Reaction ~ Days + (1 | Subject), sleepstudy)
output3 <- konfound(m3, Days, to_return = "raw_output")

m4 <- glm(outcome ~ condition, data = binary_dummy_data, family = binomial(link = "logit"))
output4 <- konfound(m4, condition, two_by_two = TRUE, n_treat = 55, to_return = "raw_output")

test_that("konfound works for linear model", {
expect_equal(output1$perc_bias_to_change, 66.57696, tolerance = .001)
library(forcats)
library(lme4)
m1 <- lm(mpg ~ wt + hp, data = mtcars)
output1 <- konfound(m1, wt, to_return = "raw_output")
expect_equal(output1$RIR_perc, 66.629, tolerance = .01)
})

# test_that("konfound works for glm, 2x2 model", {
# expect_equal(output2$percent_bias_to_change_inference, 35.357, tolerance = .001)
# })

test_that("konfound works for lme4 model", {
expect_equal(output3$perc_bias_to_change, 84.826, tolerance = .001)

m3 <- lme4::lmer(Reaction ~ Days + (1 | Subject), sleepstudy)
output3 <- konfound(m3, Days, to_return = "raw_output")
expect_equal(output3$RIR_perc, 84.826, tolerance = .001)
})

test_that("konfound works for glm, 2x2 model", {
expect_equal(output4$RIR, 15)
m4 <- glm(outcome ~ condition, data = binary_dummy_data, family = binomial(link = "logit"))
output4 <- konfound(m4, condition, two_by_two = TRUE, n_treat = 55, to_return = "raw_output")
expect_equal(output4$RIR_primary, 15)

m4 <- glm(outcome ~ condition, data = binary_dummy_data, family = binomial(link = "logit"))
output4_print <- capture.output(konfound(m4, condition, two_by_two = TRUE, n_treat = 55, to_return = "print"))
expect_true(length(output4_print) > 0)
})


test_that("konfound returns a tibble", {
m5 <- lm(mpg ~ wt + hp, data = mtcars)
output5 <- konfound(m5, wt, to_return = "table")

expect_s3_class(output5, "tbl_df")

mtcars$my_var <- runif(nrow(mtcars))
m5b <- lm(wt ~ my_var, data = mtcars)
output5b <- konfound(m5b, my_var, to_return = "table")

expect_s3_class(output5b, "tbl_df")
})

test_that("konfound glm works", {
gss_cat$married <- ifelse(gss_cat$marital == "Married", 1, 0)

m6 <- glm(married ~ age, data = gss_cat, family = binomial(link = "logit"))
m6_output <- konfound(m6, age, to_return = "raw_output")

expect_equal(as.vector(m6_output$RIR_perc), 35.357, tolerance = .001)
})
21 changes: 8 additions & 13 deletions tests/testthat/test-mkonfound.r
View file
Original file line number Diff line number Diff line change
@@ -1,17 +1,12 @@
context("Checking mkonfound")

library(lme4)
library(mice)

popmis <- popmis[1:100, ]

testmod1 <- lm(teachpop ~ texp + sex, data = popmis)
testmod2 <- lmer(teachpop ~ texp + sex + (1 | school), data = popmis)

output1 <- konfound(testmod1, texp, test_all = TRUE, to_return = "raw_output")
output2 <- konfound(testmod2, texp, test_all = TRUE, to_return = "raw_output")
test_that("mkonfound produces table", {
d <- mkonfound(mkonfound_ex, t, df)
expect_is(d, "data.frame")
})

test_that("mkonfound works for lm and lmer outout", {
expect_is(output1, "data.frame")
expect_is(output2, "data.frame")
test_that("mkonfound produces plot", {
p <- mkonfound(mkonfound_ex, t, df, return_plot = TRUE)
p
expect_s3_class(p, "ggplot")
})
184 changes: 151 additions & 33 deletions tests/testthat/test-pkonfound.R
View file
Original file line number Diff line number Diff line change
@@ -1,52 +1,170 @@
context("Checking pkonfound")

library(lme4)
library(mice)

popmis <- popmis[1:100, ]
testmod1 <- lm(teachpop ~ texp + sex, data = popmis)
testmod2 <- lmer(teachpop ~ texp + sex + (1 | school), data = popmis)

output1 <- konfound(testmod1, texp, test_all = TRUE, to_return = "raw_output")
output2 <- konfound(testmod2, texp, test_all = TRUE, to_return = "raw_output")

output3 <- pkonfound(a = 35, b = 17, c = 17, d = 38, to_return = "raw_output")

test_that("pkonfound test positive, significant coefficient works", {
expect_equal(pkonfound(2, .4, 100, 3, to_return = "raw_output")$perc_bias_to_change, tolerance = .001, 60.28953) # pct bias
expect_equal(pkonfound(2, .4, 100, 3, to_return = "raw_output")$rxcvGz, tolerance = .001, 0.568) # r-corr
expect_equal(pkonfound(2, .4, 100, 3, to_return = "raw_output")$itcvGz, tolerance = .001, 0.3224053) # impact
expect_equal(pkonfound(2, .4, 100, 3, to_return = "raw_output")$perc_bias_to_change, tolerance = .001, .603 * 100) # pct bias
expect_equal(pkonfound(2, .4, 100, 3, to_return = "raw_output")$rxcvGz, tolerance = .001, 0.566) # rxcvGz
expect_equal(pkonfound(2, .4, 100, 3, to_return = "raw_output")$itcvGz, tolerance = .001, 0.321) # itcvGz
})

test_that("pkonfound test negative, significant coefficient works", {
expect_equal(pkonfound(-2, .4, 100, 3, to_return = "raw_output")$perc_bias_to_change, tolerance = .001, 60.28953)
expect_equal(pkonfound(-2, .4, 100, 3, to_return = "raw_output")$rxcvGz, tolerance = .001, 0.568)
expect_equal(pkonfound(-2, .4, 100, 3, to_return = "raw_output")$itcvGz, tolerance = .001, -0.3224053)
expect_equal(pkonfound(-2, .4, 100, 3, to_return = "raw_output")$perc_bias_to_change, tolerance = .001, .603 * 100)
expect_equal(pkonfound(-2, .4, 100, 3, to_return = "raw_output")$rxcvGz, tolerance = .001, 0.566)
expect_equal(pkonfound(-2, .4, 100, 3, to_return = "raw_output")$itcvGz, tolerance = .001, -0.321)
})

test_that("pkonfound test positive, not significant coefficient works", {
expect_equal(pkonfound(1, .4, 100, 3, to_return = "raw_output")$perc_bias_to_change, tolerance = .001, 20.579)
expect_equal(pkonfound(1, .4, 100, 3, to_return = "raw_output")$rxcvGz, tolerance = .001, 0.248)
expect_equal(pkonfound(1, .4, 100, 3, to_return = "raw_output")$itcvGz, tolerance = .001, 0.061)
expect_equal(pkonfound(1, .4, 100, 3, to_return = "raw_output")$perc_bias_to_change, tolerance = .001, .206 * 100)
expect_equal(pkonfound(1, .4, 100, 3, to_return = "raw_output")$rxcvGz, tolerance = .001, 0.247)
expect_equal(pkonfound(1, .4, 100, 3, to_return = "raw_output")$itcvGz, tolerance = .001, 0.061)
})

test_that("pkonfound test negative, not significant coefficient works", {
expect_equal(pkonfound(-1, .4, 100, 3, to_return = "raw_output")$perc_bias_to_change, tolerance = .001, 20.579)
expect_equal(pkonfound(-1, .4, 100, 3, to_return = "raw_output")$rxcvGz, tolerance = .001, 0.248)
expect_equal(pkonfound(-1, .4, 100, 3, to_return = "raw_output")$itcvGz, tolerance = .001, -.061)
expect_equal(pkonfound(-1, .4, 100, 3, to_return = "raw_output")$perc_bias_to_change, tolerance = .001, 0.206 * 100)
expect_equal(pkonfound(-1, .4, 100, 3, to_return = "raw_output")$rxcvGz, tolerance = .001, 0.247)
expect_equal(pkonfound(-1, .4, 100, 3, to_return = "raw_output")$itcvGz, tolerance = .001, -.061)
})

test_that("pkonfound creates the threshhold plot", {
thresh_plot <- pkonfound(2, .4, 100, 3, to_return = "thresh_plot")

expect_s3_class(thresh_plot, "ggplot")

thresh_plot_null <- pkonfound(.01, .4, 100, 3, to_return = "thresh_plot")

expect_s3_class(thresh_plot_null, "ggplot")
})

test_that("pkonfound creates the correlation plot", {
corr_plot <- pkonfound(2, .4, 100, 3, to_return = "corr_plot")

expect_s3_class(corr_plot, "ggplot")

corr_plot_null <- pkonfound(.01, .4, 100, 3, to_return = "corr_plot")

expect_s3_class(corr_plot_null, "ggplot")
})

expect_output(pkonfound(2, .4, 100, 3), ".")

## PSE and COP

test_that("PSE and COE work via pkonfound", {

output4 <- pkonfound(est_eff = -.125,
std_err = .050,
n_obs = 6265,
n_covariates = 7,
sdx = .217,
sdy = .991,
R2 = .251,
eff_thr = 0,
FR2max = .61,
index = "COP",
to_return = "raw_output")

output5 <- pkonfound(est_eff = .5,
std_err = .056,
n_obs = 6174,
eff_thr = .1,
sdx = 0.22,
sdy = 1,
R2 = .3,
index = "PSE",to_return = "raw_output")

expect_equal(output4$delta_exact, tolerance = .001, 1.309) # COP - delta exact

expect_equal(output5$`correlation between X and CV conditional on Z`, tolerance = .001, 0.248) # PSE
expect_equal(output5$`correlation between Y and CV conditional on Z`, tolerance = .001, 0.372) # PSE
expect_equal(output5$`correlation between X and CV`, tolerance = .001, 0.214) # PSE
expect_equal(output5$`correlation between Y and CV`, tolerance = .001, 0.313) # PSE
})

test_that("PSE and COE work via pkonfound", {

output6 <- pkonfound(est_eff = -.125,
std_err = .050,
n_obs = 6265,
n_covariates = 7,
sdx = .217,
sdy = .991,
R2 = .251,
eff_thr = 0,
FR2max = .61,
index = "COP",
to_return = "raw_output")

expect_s3_class(output6$Figure, "ggplot")
})

test_that("pkonfound test positive, significant coefficient, small sample size works", {
expect_equal(pkonfound(2, .3, 20, 3, to_return = "raw_output")$perc_bias_to_change, tolerance = .001, 67.8282)
expect_equal(pkonfound(2, .3, 20, 3, to_return = "raw_output")$rxcvGz, tolerance = .001, 0.863)
expect_equal(pkonfound(2, .3, 20, 3, to_return = "raw_output")$itcvGz, tolerance = .001, 0.74545)
## logistic

test_that("logistic models work with pkonfound", {

output7 <- pkonfound(.273, .024, 16999, 3, n_treat = 16000, model_type = "logistic", to_return = "raw_output")$RIR_primary

expect_equal(output7, 1156)

output8 <- pkonfound(.027, .024, 16999, 3, n_treat = 16000, model_type = "logistic", to_return = "raw_output")$RIR_primary

expect_equal(output8, 803)

output8_print <- capture.output(pkonfound(.027, .024, 16999, 3, n_treat = 16000, model_type = "logistic", to_return = "print"))

expect_true(length(output8_print) > 0)

output9 <- pkonfound(.027, .024, 16999, 3, n_treat = 16000, replace = "entire",
model_type = "logistic", to_return = "raw_output")$RIR_primary

expect_equal(output9, 793)

output10 <- pkonfound(.027, .024, 16999, 3, n_treat = 16000, switch_trm = FALSE,
model_type = "logistic", to_return = "raw_output")$RIR_primary

expect_equal(output10, 52)

})

## two by two table

test_that("two by two works with pkonfound", {

output11 <- pkonfound(a = 18, b = 12, c = 12, d = 17, to_return = "raw_output")$RIR_primary
expect_equal(output11, 8)

output12 <- pkonfound(a = 18, b = 12, c = 12, d = 17, switch_trm = FALSE, to_return = "raw_output")$RIR_primary
expect_equal(output12, 5)

output13 <- pkonfound(a = 18, b = 3, c = 12, d = 1, test = "chisq", to_return = "raw_output")$RIR_primary
expect_equal(output13, 6)

output14 <- pkonfound(a = 18, b = 1, c = 12, d = 1, switch_trm = FALSE, to_return = "raw_output")$RIR_primary
expect_equal(output14, 5)
})

test_that("tkonfound for two_by_two works", {
expect_equal(output3$RIR, tolerance = .001, 14)
test_that("pkonfound printed output works for a positive case", {
outputa <- capture.output(pkonfound(2, .4, 100, 3, to_return = "print", index = "RIR"))
expect_true(length(outputa) > 0)

outputb <- capture.output(pkonfound(2, .4, 100, 3, to_return = "print", index = "IT"))
expect_true(length(outputb) > 0)

outputc <- capture.output(pkonfound(.01, .4, 100, 3, to_return = "print", index = "RIR"))
expect_true(length(outputa) > 0)

outputd <- capture.output(pkonfound(.01, .4, 100, 3, to_return = "print", index = "IT"))
expect_true(length(outputb) > 0)
})

test_that("test_all works for lm and lmer outout", {
expect_is(output1, "data.frame")
expect_is(output2, "data.frame")
test_that("pkonfound printed output works for a negative case", {
output <- capture.output(pkonfound(-2.2, .65, 200, 3, to_return = "print", index = "RIR"))
expect_true(length(output) > 0)

output <- capture.output(pkonfound(-2.2, .65, 200, 3, to_return = "print", index = "IT"))
expect_true(length(output) > 0)

output <- capture.output(pkonfound(-.01, .65, 200, 3, to_return = "print", index = "RIR"))
expect_true(length(output) > 0)

output <- capture.output(pkonfound(-.01, .65, 200, 3, to_return = "print", index = "IT"))
expect_true(length(output) > 0)
})
51 changes: 51 additions & 0 deletions tests/testthat/test-tkonfound.R
View file
Original file line number Diff line number Diff line change
@@ -0,0 +1,51 @@
context("Checking tkonfound")

test_that("tkonfound creates the correlation plot", {

tkonfound_fig <- tkonfound_fig(a = 35, b = 17, c = 17, d = 38)[[3]]

expect_s3_class(tkonfound_fig, "ggplot")
})

test_that("tkonfound creates the correlation plot when switch_trm = FALSE", {

tkonfound_fig2 <- tkonfound_fig(a = 35, b = 17, c = 17, d = 38, switch_trm = FALSE)[[3]]

expect_s3_class(tkonfound_fig2, "ggplot")
})

test_that("tkonfound creates the correlation plot with test = chisq", {

tkonfound_fig3 <- tkonfound_fig(a = 35, b = 17, c = 17, d = 38, switch_trm = TRUE, test = "chisq")[[3]]

expect_s3_class(tkonfound_fig3, "ggplot")
})

test_that("tkonfound creates the correlation plot with test = chisq and replace = entire", {

tkonfound_fig4 <- tkonfound_fig(a = 35, b = 17, c = 17, d = 38, switch_trm = TRUE, test = "chisq", replace = "entire")[[3]]

expect_s3_class(tkonfound_fig4, "ggplot")
})

test_that("tkonfound creates the correlation plot when treatment is small", {

tkonfound_fig5 <- tkonfound_fig(a = 35, b = 17, c = 17, d = 38, switch_trm = TRUE, test = "chisq", replace = "entire")[[3]]

expect_s3_class(tkonfound_fig5, "ggplot")
})

test_that("tkonfound for two_by_two works for raw output", {
output3 <- pkonfound(a = 35, b = 17, c = 17, d = 38, to_return = "raw_output")

expect_equal(output3$RIR_primary, tolerance = .001, 14)

output3_null <- pkonfound(a = 5, b = 17, c = 17, d = 10, to_return = "raw_output")

expect_equal(output3_null$RIR_primary, tolerance = .001, 6)

pkonfound_two_by_two <- capture.output(pkonfound(a = 35, b = 17, c = 17, d = 38, to_return = "print")) # printed output works

expect_true(length(pkonfound_two_by_two) > 0)

})

0 comments on commit 04fbc43

Please sign in to comment.