Package 'cardx'

Description

Function takes a regression model object and calculates Wald statistical test using aod::wald.test().

Usage

ard_aod_wald_test(
  x,
  tidy_fun = broom.helpers::tidy_with_broom_or_parameters,
  ...
)

Arguments

Value

data frame

Examples

lm(AGE ~ ARM, data = cards::ADSL) |>
  ard_aod_wald_test()

Description

Add variable attributes to an ARD data frame.

• The label attribute will be added for all columns, and when no label is specified and no label has been set for a column using the ⁠label=⁠ argument, the column name will be placed in the label statistic.

• The class attribute will also be returned for all columns.

• Any other attribute returned by attributes() will also be added, e.g. factor levels.

Usage

## S3 method for class 'survey.design'
ard_attributes(data, variables = everything(), label = NULL, ...)

Arguments

Value

an ARD data frame of class 'card'

Examples

data(api, package = "survey")
dclus1 <- survey::svydesign(id = ~dnum, weights = ~pw, data = apiclus1, fpc = ~fpc)

ard_attributes(
  data = dclus1,
  variables = c(sname, dname),
  label = list(sname = "School Name", dname = "District Name")
)

Description

Function takes a regression model object and calculated ANOVA using car::Anova().

Usage

ard_car_anova(x, ...)

Arguments

Value

data frame

Examples

lm(AGE ~ ARM, data = cards::ADSL) |>
  ard_car_anova()

glm(vs ~ factor(cyl) + factor(am), data = mtcars, family = binomial) |>
  ard_car_anova(test.statistic = "Wald")

Description

Function takes a regression model object and returns the variance inflation factor (VIF) using car::vif() and converts it to a ARD structure

Usage

ard_car_vif(x, ...)

Arguments

Value

data frame

Examples

lm(AGE ~ ARM + SEX, data = cards::ADSL) |>
  ard_car_vif()

Description

Calculate confidence intervals for proportions.

Usage

ard_categorical_ci(data, ...)

## S3 method for class 'data.frame'
ard_categorical_ci(
  data,
  variables,
  by = dplyr::group_vars(data),
  method = c("waldcc", "wald", "clopper-pearson", "wilson", "wilsoncc", "strat_wilson",
    "strat_wilsoncc", "agresti-coull", "jeffreys"),
  conf.level = 0.95,
  value = list(where(is_binary) ~ 1L, where(is.logical) ~ TRUE),
  strata = NULL,
  weights = NULL,
  max.iterations = 10,
  ...
)

Arguments

Value

an ARD data frame

Examples

# compute CI for binary variables
ard_categorical_ci(mtcars, variables = c(vs, am), method = "wilson")

# compute CIs for each level of a categorical variable
ard_categorical_ci(mtcars, variables = cyl, method = "jeffreys")

Description

Confidence intervals for categorical variables calculated via survey::svyciprop().

Usage

## S3 method for class 'survey.design'
ard_categorical_ci(
  data,
  variables,
  by = NULL,
  method = c("logit", "likelihood", "asin", "beta", "mean", "xlogit"),
  conf.level = 0.95,
  value = list(where(is_binary) ~ 1L, where(is.logical) ~ TRUE),
  df = survey::degf(data),
  ...
)

Arguments

Value

ARD data frame

Examples

data(api, package = "survey")
dclus1 <- survey::svydesign(id = ~dnum, weights = ~pw, data = apiclus1, fpc = ~fpc)

ard_categorical_ci(dclus1, variables = sch.wide)
ard_categorical_ci(dclus1, variables = sch.wide, value = sch.wide ~ "Yes", method = "xlogit")

Description

Compute tabulations on survey-weighted data.

The counts and proportion ("N", "n", "p") are calculated using survey::svytable(), and the standard errors and design effect ("p.std.error", "deff") are calculated using survey::svymean().

The unweighted statistics are calculated with cards::ard_categorical.data.frame().

Usage

## S3 method for class 'survey.design'
ard_categorical(
  data,
  variables,
  by = NULL,
  statistic = everything() ~ c("n", "N", "p", "p.std.error", "deff", "n_unweighted",
    "N_unweighted", "p_unweighted"),
  denominator = c("column", "row", "cell"),
  fmt_fn = NULL,
  stat_label = everything() ~ list(p = "%", p.std.error = "SE(%)", deff =
    "Design Effect", n_unweighted = "Unweighted n", N_unweighted = "Unweighted N",
    p_unweighted = "Unweighted %"),
  ...
)

Arguments

Value

an ARD data frame of class 'card'

Examples

svy_titanic <- survey::svydesign(~1, data = as.data.frame(Titanic), weights = ~Freq)

ard_categorical(svy_titanic, variables = c(Class, Age), by = Survived)

Description

One-sample confidence intervals for continuous variable means and medians.

Usage

ard_continuous_ci(data, ...)

## S3 method for class 'data.frame'
ard_continuous_ci(
  data,
  variables,
  by = dplyr::group_vars(data),
  conf.level = 0.95,
  method = c("t.test", "wilcox.test"),
  ...
)

Arguments

Value

ARD data frame

Examples

ard_continuous_ci(mtcars, variables = c(mpg, hp), method = "wilcox.test")
ard_continuous_ci(mtcars, variables = mpg, by = am, method = "t.test")

Description

One-sample confidence intervals for continuous variables' means and medians. Confidence limits are calculated with survey::svymean() and survey::svyquantile().

Usage

## S3 method for class 'survey.design'
ard_continuous_ci(
  data,
  variables,
  by = NULL,
  method = c("svymean", "svymedian.mean", "svymedian.beta", "svymedian.xlogit",
    "svymedian.asin", "svymedian.score"),
  conf.level = 0.95,
  df = survey::degf(data),
  ...
)

Arguments

Value

ARD data frame

Examples

data(api, package = "survey")
dclus1 <- survey::svydesign(id = ~dnum, weights = ~pw, data = apiclus1, fpc = ~fpc)

ard_continuous_ci(dclus1, variables = api00)
ard_continuous_ci(dclus1, variables = api00, method = "svymedian.xlogit")

Description

Returns an ARD of weighted statistics using the {survey} package.

Usage

## S3 method for class 'survey.design'
ard_continuous(
  data,
  variables,
  by = NULL,
  statistic = everything() ~ c("median", "p25", "p75"),
  fmt_fn = NULL,
  stat_label = NULL,
  ...
)

Arguments

Value

an ARD data frame of class 'card'

statistic argument

The following statistics are available: 'mean', 'median', 'min', 'max', 'sum', 'var', 'sd', 'mean.std.error', 'deff', 'p##', where 'p##' is are the percentiles and ⁠##⁠ is an integer between 0 and 100.

Examples

data(api, package = "survey")
dclus1 <- survey::svydesign(id = ~dnum, weights = ~pw, data = apiclus1, fpc = ~fpc)

ard_continuous(
  data = dclus1,
  variables = api00,
  by = stype
)

Description

Compute Analysis Results Data (ARD) for dichotomous summary statistics.

Usage

## S3 method for class 'survey.design'
ard_dichotomous(
  data,
  variables,
  by = NULL,
  value = cards::maximum_variable_value(data$variables[variables]),
  statistic = everything() ~ c("n", "N", "p", "p.std.error", "deff", "n_unweighted",
    "N_unweighted", "p_unweighted"),
  denominator = c("column", "row", "cell"),
  fmt_fn = NULL,
  stat_label = everything() ~ list(p = "%", p.std.error = "SE(%)", deff =
    "Design Effect", n_unweighted = "Unweighted n", N_unweighted = "Unweighted N",
    p_unweighted = "Unweighted %"),
  ...
)

Arguments

Value

an ARD data frame of class 'card'

Examples

survey::svydesign(ids = ~1, data = mtcars, weights = ~1) |>
  ard_dichotomous(by = vs, variables = c(cyl, am), value = list(cyl = 4))

Description

Analysis results data for paired and non-paired Cohen's D Effect Size Test using effectsize::cohens_d().

Usage

ard_effectsize_cohens_d(data, by, variables, conf.level = 0.95, ...)

ard_effectsize_paired_cohens_d(data, by, variables, id, conf.level = 0.95, ...)

Arguments

Details

For the ard_effectsize_cohens_d() function, the data is expected to be one row per subject. The data is passed as effectsize::cohens_d(data[[variable]]~data[[by]], data, paired = FALSE, ...).

For the ard_effectsize_paired_cohens_d() function, the data is expected to be one row per subject per by level. Before the effect size is calculated, the data are reshaped to a wide format to be one row per subject. The data are then passed as ⁠effectsize::cohens_d(x = data_wide[[<by level 1>]], y = data_wide[[<by level 2>]], paired = TRUE, ...)⁠.

Value

ARD data frame

Examples

cards::ADSL |>
  dplyr::filter(ARM %in% c("Placebo", "Xanomeline High Dose")) |>
  ard_effectsize_cohens_d(by = ARM, variables = AGE)

# constructing a paired data set,
# where patients receive both treatments
cards::ADSL[c("ARM", "AGE")] |>
  dplyr::filter(ARM %in% c("Placebo", "Xanomeline High Dose")) |>
  dplyr::mutate(.by = ARM, USUBJID = dplyr::row_number()) |>
  dplyr::arrange(USUBJID, ARM) |>
  dplyr::group_by(USUBJID) |>
  dplyr::filter(dplyr::n() > 1) |>
  ard_effectsize_paired_cohens_d(by = ARM, variables = AGE, id = USUBJID)

Description

Analysis results data for paired and non-paired Hedge's G Effect Size Test using effectsize::hedges_g().

Usage

ard_effectsize_hedges_g(data, by, variables, conf.level = 0.95, ...)

ard_effectsize_paired_hedges_g(data, by, variables, id, conf.level = 0.95, ...)

Arguments

Details

For the ard_effectsize_hedges_g() function, the data is expected to be one row per subject. The data is passed as effectsize::hedges_g(data[[variable]]~data[[by]], data, paired = FALSE, ...).

For the ard_effectsize_paired_hedges_g() function, the data is expected to be one row per subject per by level. Before the effect size is calculated, the data are reshaped to a wide format to be one row per subject. The data are then passed as ⁠effectsize::hedges_g(x = data_wide[[<by level 1>]], y = data_wide[[<by level 2>]], paired = TRUE, ...)⁠.

Value

ARD data frame

Examples

cards::ADSL |>
  dplyr::filter(ARM %in% c("Placebo", "Xanomeline High Dose")) |>
  ard_effectsize_hedges_g(by = ARM, variables = AGE)

# constructing a paired data set,
# where patients receive both treatments
cards::ADSL[c("ARM", "AGE")] |>
  dplyr::filter(ARM %in% c("Placebo", "Xanomeline High Dose")) |>
  dplyr::mutate(.by = ARM, USUBJID = dplyr::row_number()) |>
  dplyr::arrange(USUBJID, ARM) |>
  dplyr::group_by(USUBJID) |>
  dplyr::filter(dplyr::n() > 1) |>
  ard_effectsize_paired_hedges_g(by = ARM, variables = AGE, id = USUBJID)

Description

This function calculates least-squares mean differences using the 'emmeans' package using the following

emmeans::emmeans(object = <regression model>, specs = ~ <primary covariate>) |>
  emmeans::contrast(method = "pairwise") |>
  summary(infer = TRUE, level = <confidence level>)

The arguments data, formula, method, method.args, package are used to construct the regression model via cardx::construct_model().

Usage

ard_emmeans_mean_difference(
  data,
  formula,
  method,
  method.args = list(),
  package = "base",
  response_type = c("continuous", "dichotomous"),
  conf.level = 0.95,
  primary_covariate = getElement(attr(stats::terms(formula), "term.labels"), 1L)
)

Arguments

Value

ARD data frame

Examples

ard_emmeans_mean_difference(
  data = mtcars,
  formula = mpg ~ am + cyl,
  method = "lm"
)

ard_emmeans_mean_difference(
  data = mtcars,
  formula = vs ~ am + mpg,
  method = "glm",
  method.args = list(family = binomial),
  response_type = "dichotomous"
)

Description

Compute Analysis Results Data (ARD) for statistics related to data missingness for survey objects

Usage

## S3 method for class 'survey.design'
ard_missing(
  data,
  variables,
  by = NULL,
  statistic = everything() ~ c("N_obs", "N_miss", "N_nonmiss", "p_miss", "p_nonmiss",
    "N_obs_unweighted", "N_miss_unweighted", "N_nonmiss_unweighted", "p_miss_unweighted",
    "p_nonmiss_unweighted"),
  fmt_fn = NULL,
  stat_label = everything() ~ list(N_obs = "Total N", N_miss = "N Missing", N_nonmiss =
    "N not Missing", p_miss = "% Missing", p_nonmiss = "% not Missing",
    N_obs_unweighted = "Total N (unweighted)", N_miss_unweighted =
    "N Missing (unweighted)", N_nonmiss_unweighted = "N not Missing (unweighted)",
    p_miss_unweighted = "% Missing (unweighted)", p_nonmiss_unweighted =
    "% not Missing (unweighted)"),
  ...
)

Arguments

Value

an ARD data frame of class 'card'

Examples

svy_titanic <- survey::svydesign(~1, data = as.data.frame(Titanic), weights = ~Freq)

ard_missing(svy_titanic, variables = c(Class, Age), by = Survived)

Description

Function takes a regression model object and converts it to a ARD structure using the broom.helpers package.

Usage

ard_regression(x, ...)

## Default S3 method:
ard_regression(x, tidy_fun = broom.helpers::tidy_with_broom_or_parameters, ...)

Arguments

Value

data frame

Examples

lm(AGE ~ ARM, data = cards::ADSL) |>
  ard_regression(add_estimate_to_reference_rows = TRUE)

Description

A function that takes a regression model and provides basic statistics in an ARD structure. The default output is simpler than ard_regression(). The function primarily matches regression terms to underlying variable names and levels. The default arguments used are

broom.helpers::tidy_plus_plus(
  add_reference_rows = FALSE,
  add_estimate_to_reference_rows = FALSE,
  add_n = FALSE,
  intercept = FALSE
)

Usage

ard_regression_basic(
  x,
  tidy_fun = broom.helpers::tidy_with_broom_or_parameters,
  stats_to_remove = c("term", "var_type", "var_label", "var_class", "label",
    "contrasts_type", "contrasts", "var_nlevels"),
  ...
)

Arguments

Value

data frame

Examples

lm(AGE ~ ARM, data = cards::ADSL) |>
  ard_regression_basic()

Description

Standardized mean difference calculated via smd::smd() with na.rm = TRUE. Additionally, this function add a confidence interval to the SMD when std.error=TRUE, which the original smd::smd() does not include.

Usage

ard_smd_smd(data, by, variables, std.error = TRUE, conf.level = 0.95, ...)

Arguments

Value

ARD data frame

Examples

ard_smd_smd(cards::ADSL, by = SEX, variables = AGE)
ard_smd_smd(cards::ADSL, by = SEX, variables = AGEGR1)

Description

Prepare ANOVA results from the stats::anova() function. Users may pass a pre-calculated stats::anova() object or a list of formulas. In the latter case, the models will be constructed using the information passed and models will be passed to stats::anova().

Usage

ard_stats_anova(x, ...)

## S3 method for class 'anova'
ard_stats_anova(x, method_text = "ANOVA results from `stats::anova()`", ...)

## S3 method for class 'data.frame'
ard_stats_anova(
  x,
  formulas,
  method,
  method.args = list(),
  package = "base",
  method_text = "ANOVA results from `stats::anova()`",
  ...
)

Arguments

Details

When a list of formulas is supplied to ard_stats_anova(), these formulas along with information from other arguments, are used to construct models and pass those models to stats::anova().

The models are constructed using rlang::exec(), which is similar to do.call().

rlang::exec(.fn = method, formula = formula, data = data, !!!method.args)

The above function is executed in withr::with_namespace(package), which allows for the use of ard_stats_anova(method) from packages, e.g. package = 'lme4' must be specified when method = 'glmer'. See example below.

Value

ARD data frame

Examples

anova(
  lm(mpg ~ am, mtcars),
  lm(mpg ~ am + hp, mtcars)
) |>
  ard_stats_anova()

ard_stats_anova(
  x = mtcars,
  formulas = list(am ~ mpg, am ~ mpg + hp),
  method = "glm",
  method.args = list(family = binomial)
)

ard_stats_anova(
  x = mtcars,
  formulas = list(am ~ 1 + (1 | vs), am ~ mpg + (1 | vs)),
  method = "glmer",
  method.args = list(family = binomial),
  package = "lme4"
)

Description

Analysis results data for Analysis of Variance. Calculated with stats::aov()

Usage

ard_stats_aov(formula, data, ...)

Arguments

Value

ARD data frame

Examples

ard_stats_aov(AGE ~ ARM, data = cards::ADSL)

Description

Analysis results data for Pearson's Chi-squared Test. Calculated with chisq.test(x = data[[variable]], y = data[[by]], ...)

Usage

ard_stats_chisq_test(data, by, variables, ...)

Arguments

Value

ARD data frame

Examples

cards::ADSL |>
  ard_stats_chisq_test(by = "ARM", variables = "AGEGR1")

Description

Analysis results data for Fisher's Exact Test. Calculated with fisher.test(x = data[[variable]], y = data[[by]], ...)

Usage

ard_stats_fisher_test(data, by, variables, conf.level = 0.95, ...)

Arguments

Value

ARD data frame

Examples

cards::ADSL[1:30, ] |>
  ard_stats_fisher_test(by = "ARM", variables = "AGEGR1")

Description

Analysis results data for Kruskal-Wallis Rank Sum Test.

Calculated with kruskal.test(data[[variable]], data[[by]], ...)

Usage

ard_stats_kruskal_test(data, by, variables)

Arguments

Value

ARD data frame

Examples

cards::ADSL |>
  ard_stats_kruskal_test(by = "ARM", variables = "AGE")

Description

Analysis results data for McNemar's statistical test. We have two functions depending on the structure of the data.

• ard_stats_mcnemar_test() is the structure expected by stats::mcnemar.test()

• ard_stats_mcnemar_test_long() is one row per ID per group

Usage

ard_stats_mcnemar_test(data, by, variables, ...)

ard_stats_mcnemar_test_long(data, by, variables, id, ...)

Arguments

Details

For the ard_stats_mcnemar_test() function, the data is expected to be one row per subject. The data is passed as stats::mcnemar.test(x = data[[variable]], y = data[[by]], ...). Please use table(x = data[[variable]], y = data[[by]]) to check the contingency table.

Value

ARD data frame

Examples

cards::ADSL |>
  ard_stats_mcnemar_test(by = "SEX", variables = "EFFFL")

set.seed(1234)
cards::ADSL[c("USUBJID", "TRT01P")] |>
  dplyr::mutate(TYPE = "PLANNED") |>
  dplyr::rename(TRT01 = TRT01P) %>%
  dplyr::bind_rows(dplyr::mutate(., TYPE = "ACTUAL", TRT01 = sample(TRT01))) |>
  ard_stats_mcnemar_test_long(
    by = TYPE,
    variable = TRT01,
    id = USUBJID
  )

Description

Analysis results data for Mood two sample test of scale. Note this not to be confused with the Brown-Mood test of medians.

Usage

ard_stats_mood_test(data, by, variables, ...)

Arguments

Details

For the ard_stats_mood_test() function, the data is expected to be one row per subject. The data is passed as mood.test(data[[variable]] ~ data[[by]], ...).

Value

ARD data frame

Examples

cards::ADSL |>
  ard_stats_mood_test(by = "SEX", variables = "AGE")

Description

Analysis results data for Testing Equal Means in a One-Way Layout. calculated with oneway.test()

Usage

ard_stats_oneway_test(formula, data, ...)

Arguments

Value

ARD data frame

Examples

ard_stats_oneway_test(AGE ~ ARM, data = cards::ADSL)

Description

Analysis results data for exact tests of a simple null hypothesis about the rate parameter in Poisson distribution, or the comparison of two rate parameters.

Usage

ard_stats_poisson_test(
  data,
  variables,
  na.rm = TRUE,
  by = NULL,
  conf.level = 0.95,
  ...
)

Arguments

Details

• For the ard_stats_poisson_test() function, the data is expected to be one row per subject.

• If by is not specified, an exact Poisson test of the rate parameter will be performed. Otherwise, a Poisson comparison of two rate parameters will be performed on the levels of by. If by has more than 2 levels, an error will occur.

Value

an ARD data frame of class 'card'

Examples

# Exact test of rate parameter against null hypothesis
cards::ADTTE |>
  ard_stats_poisson_test(variables = c(CNSR, AVAL))

# Comparison test of ratio of 2 rate parameters against null hypothesis
cards::ADTTE |>
  dplyr::filter(TRTA %in% c("Placebo", "Xanomeline High Dose")) |>
  ard_stats_poisson_test(by = TRTA, variables = c(CNSR, AVAL))

Description

Analysis results data for a 2-sample test or proportions using stats::prop.test().

Usage

ard_stats_prop_test(data, by, variables, conf.level = 0.95, ...)

Arguments

Value

ARD data frame

Examples

mtcars |>
  ard_stats_prop_test(by = vs, variables = am)

Description

Analysis results data for paired and non-paired t-tests.

Usage

ard_stats_t_test(data, variables, by = NULL, conf.level = 0.95, ...)

ard_stats_paired_t_test(data, by, variables, id, conf.level = 0.95, ...)

Arguments

Details

For the ard_stats_t_test() function, the data is expected to be one row per subject. The data is passed as t.test(data[[variable]] ~ data[[by]], paired = FALSE, ...).

For the ard_stats_paired_t_test() function, the data is expected to be one row per subject per by level. Before the t-test is calculated, the data are reshaped to a wide format to be one row per subject. The data are then passed as ⁠t.test(x = data_wide[[<by level 1>]], y = data_wide[[<by level 2>]], paired = TRUE, ...)⁠.

Value

ARD data frame

Examples

cards::ADSL |>
  dplyr::filter(ARM %in% c("Placebo", "Xanomeline High Dose")) |>
  ard_stats_t_test(by = ARM, variables = c(AGE, BMIBL))

# constructing a paired data set,
# where patients receive both treatments
cards::ADSL[c("ARM", "AGE")] |>
  dplyr::filter(ARM %in% c("Placebo", "Xanomeline High Dose")) |>
  dplyr::mutate(.by = ARM, USUBJID = dplyr::row_number()) |>
  dplyr::arrange(USUBJID, ARM) |>
  ard_stats_paired_t_test(by = ARM, variables = AGE, id = USUBJID)

Description

Analysis results data for one-sample t-tests. Result may be stratified by including the by argument.

Usage

ard_stats_t_test_onesample(
  data,
  variables,
  by = dplyr::group_vars(data),
  conf.level = 0.95,
  ...
)

Arguments

Value

ARD data frame

Examples

cards::ADSL |>
  ard_stats_t_test_onesample(by = ARM, variables = AGE)

Description

Analysis results data for paired and non-paired Wilcoxon Rank-Sum tests.

Usage

ard_stats_wilcox_test(data, variables, by = NULL, conf.level = 0.95, ...)

ard_stats_paired_wilcox_test(data, by, variables, id, conf.level = 0.95, ...)

Arguments

Details

For the ard_stats_wilcox_test() function, the data is expected to be one row per subject. The data is passed as wilcox.test(data[[variable]] ~ data[[by]], paired = FALSE, ...).

For the ard_stats_paired_wilcox_test() function, the data is expected to be one row per subject per by level. Before the test is calculated, the data are reshaped to a wide format to be one row per subject. The data are then passed as ⁠wilcox.test(x = data_wide[[<by level 1>]], y = data_wide[[<by level 2>]], paired = TRUE, ...)⁠.

Value

ARD data frame

Examples

cards::ADSL |>
  dplyr::filter(ARM %in% c("Placebo", "Xanomeline High Dose")) |>
  ard_stats_wilcox_test(by = "ARM", variables = "AGE")

# constructing a paired data set,
# where patients receive both treatments
cards::ADSL[c("ARM", "AGE")] |>
  dplyr::filter(ARM %in% c("Placebo", "Xanomeline High Dose")) |>
  dplyr::mutate(.by = ARM, USUBJID = dplyr::row_number()) |>
  dplyr::arrange(USUBJID, ARM) |>
  ard_stats_paired_wilcox_test(by = ARM, variables = AGE, id = USUBJID)

Description

Analysis results data for one-sample Wilcox Rank-sum. Result may be stratified by including the by argument.

Usage

ard_stats_wilcox_test_onesample(
  data,
  variables,
  by = dplyr::group_vars(data),
  conf.level = 0.95,
  ...
)

Arguments

Value

ARD data frame

Examples

cards::ADSL |>
  ard_stats_wilcox_test_onesample(by = ARM, variables = AGE)

Description

Analysis results data for survey Chi-Square test using survey::svychisq(). Only two-way comparisons are supported.

Usage

ard_survey_svychisq(data, by, variables, statistic = "F", ...)

Arguments

Value

ARD data frame

Examples

data(api, package = "survey")
dclus1 <- survey::svydesign(id = ~dnum, weights = ~pw, data = apiclus1, fpc = ~fpc)

ard_survey_svychisq(dclus1, variables = sch.wide, by = comp.imp, statistic = "F")

Description

Analysis results data for survey wilcox test using survey::svyranktest().

Usage

ard_survey_svyranktest(data, by, variables, test, ...)

Arguments

Value

ARD data frame

Examples

data(api, package = "survey")
dclus2 <- survey::svydesign(id = ~ dnum + snum, fpc = ~ fpc1 + fpc2, data = apiclus2)

ard_survey_svyranktest(dclus2, variables = enroll, by = comp.imp, test = "wilcoxon")
ard_survey_svyranktest(dclus2, variables = enroll, by = comp.imp, test = "vanderWaerden")
ard_survey_svyranktest(dclus2, variables = enroll, by = comp.imp, test = "median")
ard_survey_svyranktest(dclus2, variables = enroll, by = comp.imp, test = "KruskalWallis")

Description

Analysis results data for survey t-test using survey::svyttest().

Usage

ard_survey_svyttest(data, by, variables, conf.level = 0.95, ...)

Arguments

Value

ARD data frame

Examples

data(api, package = "survey")
dclus2 <- survey::svydesign(id = ~ dnum + snum, fpc = ~ fpc1 + fpc2, data = apiclus2)

ard_survey_svyttest(dclus2, variables = enroll, by = comp.imp, conf.level = 0.9)

Description

Analysis results data for comparison of survival using survival::survdiff().

Usage

ard_survival_survdiff(formula, data, rho = 0, ...)

Arguments

Value

an ARD data frame of class 'card'

Examples

library(survival)
library(ggsurvfit)

ard_survival_survdiff(Surv_CNSR(AVAL, CNSR) ~ TRTA, data = cards::ADTTE)

Description

Analysis results data for survival quantiles and x-year survival estimates, extracted from a survival::survfit() model.

Usage

ard_survival_survfit(x, times = NULL, probs = NULL, type = NULL)

Arguments

Details

• Only one of either the times or probs parameters can be specified.

• Times should be provided using the same scale as the time variable used to fit the provided survival fit model.

Value

an ARD data frame of class 'card'

Examples

library(survival)
library(ggsurvfit)

survfit(Surv_CNSR(AVAL, CNSR) ~ TRTA, cards::ADTTE) |>
  ard_survival_survfit(times = c(60, 180))

survfit(Surv_CNSR(AVAL, CNSR) ~ TRTA, cards::ADTTE) |>
  ard_survival_survfit(probs = c(0.25, 0.5, 0.75))

# Competing Risks Example ---------------------------
set.seed(1)
ADTTE_MS <- cards::ADTTE %>%
  dplyr::mutate(
    CNSR = dplyr::case_when(
      CNSR == 0 ~ "censor",
      runif(dplyr::n()) < 0.5 ~ "death from cancer",
      TRUE ~ "death other causes"
    ) %>% factor()
  )

survfit(Surv(AVAL, CNSR) ~ TRTA, data = ADTTE_MS) %>%
  ard_survival_survfit(times = c(60, 180))

Description

Calculate differences in the Kaplan-Meier estimator of survival using the results from survival::survfit().

Usage

ard_survival_survfit_diff(x, times, conf.level = 0.95)

Arguments

Value

an ARD data frame of class 'card'

Examples

library(ggsurvfit)
library(survival)

survfit(Surv_CNSR() ~ TRTA, data = cards::ADTTE) |>
  ard_survival_survfit_diff(times = c(25, 50))

Description

Returns the total N for a survey object. The placeholder variable name returned in the object is "..ard_total_n.."

Usage

## S3 method for class 'survey.design'
ard_total_n(data, ...)

Arguments

Value

an ARD data frame of class 'card'

Examples

svy_titanic <- survey::svydesign(~1, data = as.data.frame(Titanic), weights = ~Freq)

ard_total_n(svy_titanic)

Description

These functions help construct calls to various types of models.

Usage

construct_model(data, ...)

## S3 method for class 'data.frame'
construct_model(
  data,
  formula,
  method,
  method.args = list(),
  package = "base",
  env = caller_env(),
  ...
)

## S3 method for class 'survey.design'
construct_model(
  data,
  formula,
  method,
  method.args = list(),
  package = "survey",
  env = caller_env(),
  ...
)

reformulate2(
  termlabels,
  response = NULL,
  intercept = TRUE,
  env = parent.frame(),
  pattern_term = NULL,
  pattern_response = NULL
)

bt(x, pattern = NULL)

bt_strip(x)

Arguments

Details

• construct_model(): Builds models of the form ⁠method(data = data, formula = formula, method.args!!!)⁠. If the package argument is specified, that package is temporarily attached when the model is evaluated.

• reformulate2(): This is a copy of reformulate() except that variable names that contain a space are wrapped in backticks.

• bt(): Adds backticks to a character vector.

• bt_strip(): Removes backticks from a string if it begins and ends with a backtick.

Value

depends on the calling function

Examples

construct_model(
  data = mtcars,
  formula = am ~ mpg + (1 | vs),
  method = "glmer",
  method.args = list(family = binomial),
  package = "lme4"
) |>
  broom.mixed::tidy()

construct_model(
  data = mtcars |> dplyr::rename(`M P G` = mpg),
  formula = reformulate2(c("M P G", "cyl"), response = "hp"),
  method = "lm"
) |>
  ard_regression() |>
  dplyr::filter(stat_name %in% c("term", "estimate", "p.value"))

Description

Functions to calculate different proportion confidence intervals for use in ard_proportion().

Usage

proportion_ci_wald(x, conf.level = 0.95, correct = FALSE)

proportion_ci_wilson(x, conf.level = 0.95, correct = FALSE)

proportion_ci_clopper_pearson(x, conf.level = 0.95)

proportion_ci_agresti_coull(x, conf.level = 0.95)

proportion_ci_jeffreys(x, conf.level = 0.95)

proportion_ci_strat_wilson(
  x,
  strata,
  weights = NULL,
  conf.level = 0.95,
  max.iterations = 10L,
  correct = FALSE
)

is_binary(x)

Arguments

Value

Confidence interval of a proportion.

Functions

• proportion_ci_wald(): Calculates the Wald interval by following the usual textbook definition for a single proportion confidence interval using the normal approximation.

  $\hat{p} \pm z_{\alpha/2} \sqrt{\frac{\hat{p}(1 - \hat{p})}{n}}$

• proportion_ci_wilson(): Calculates the Wilson interval by calling stats::prop.test(). Also referred to as Wilson score interval.

  $\frac{\hat{p} + \frac{z^2_{\alpha/2}}{2n} \pm z_{\alpha/2} \sqrt{\frac{\hat{p}(1 - \hat{p})}{n} + \frac{z^2_{\alpha/2}}{4n^2}}}{1 + \frac{z^2_{\alpha/2}}{n}}$

• proportion_ci_clopper_pearson(): Calculates the Clopper-Pearson interval by calling stats::binom.test(). Also referred to as the exact method.

  $\left( \frac{k}{n} \pm z_{\alpha/2} \sqrt{\frac{\frac{k}{n}(1-\frac{k}{n})}{n} + \frac{z^2_{\alpha/2}}{4n^2}} \right) / \left( 1 + \frac{z^2_{\alpha/2}}{n} \right)$

• proportion_ci_agresti_coull(): Calculates the Agresti-Coull interval (created by ⁠Alan Agresti⁠ and ⁠Brent Coull⁠) by (for 95% CI) adding two successes and two failures to the data and then using the Wald formula to construct a CI.

  $\left( \frac{\tilde{p} + z^2_{\alpha/2}/2}{n + z^2_{\alpha/2}} \pm z_{\alpha/2} \sqrt{\frac{\tilde{p}(1 - \tilde{p})}{n} + \frac{z^2_{\alpha/2}}{4n^2}} \right)$

• proportion_ci_jeffreys(): Calculates the Jeffreys interval, an equal-tailed interval based on the non-informative Jeffreys prior for a binomial proportion.

  $\left( \text{Beta}\left(\frac{k}{2} + \frac{1}{2}, \frac{n - k}{2} + \frac{1}{2}\right)_\alpha, \text{Beta}\left(\frac{k}{2} + \frac{1}{2}, \frac{n - k}{2} + \frac{1}{2}\right)_{1-\alpha} \right)$

• proportion_ci_strat_wilson(): Calculates the stratified Wilson confidence interval for unequal proportions as described in Xin YA, Su XG. Stratified Wilson and Newcombe confidence intervals for multiple binomial proportions. Statistics in Biopharmaceutical Research. 2010;2(3).

  $\frac{\hat{p}_j + \frac{z^2_{\alpha/2}}{2n_j} \pm z_{\alpha/2} \sqrt{\frac{\hat{p}_j(1 - \hat{p}_j)}{n_j} + \frac{z^2_{\alpha/2}}{4n_j^2}}}{1 + \frac{z^2_{\alpha/2}}{n_j}}$

• is_binary(): Helper to determine if vector is binary (logical or 0/1)

Examples

x <- c(
  TRUE, TRUE, TRUE, TRUE, TRUE,
  FALSE, FALSE, FALSE, FALSE, FALSE
)

proportion_ci_wald(x, conf.level = 0.9)
proportion_ci_wilson(x, correct = TRUE)
proportion_ci_clopper_pearson(x)
proportion_ci_agresti_coull(x)
proportion_ci_jeffreys(x)

# Stratified Wilson confidence interval with unequal probabilities

set.seed(1)
rsp <- sample(c(TRUE, FALSE), 100, TRUE)
strata_data <- data.frame(
  "f1" = sample(c("a", "b"), 100, TRUE),
  "f2" = sample(c("x", "y", "z"), 100, TRUE),
  stringsAsFactors = TRUE
)
strata <- interaction(strata_data)
n_strata <- ncol(table(rsp, strata)) # Number of strata

proportion_ci_strat_wilson(
  x = rsp, strata = strata,
  conf.level = 0.90
)

# Not automatic setting of weights
proportion_ci_strat_wilson(
  x = rsp, strata = strata,
  weights = rep(1 / n_strata, n_strata),
  conf.level = 0.90
)