| Type: | Package |
| Title: | Categorical Data Analysis |
| Version: | 0.1.4 |
| Author: | Nick Williams |
| Maintainer: | Nick Williams <ntwilliams.personal@gmail.com> |
| Description: | Includes wrapper functions around existing functions for the analysis of categorical data and introduces functions for calculating risk differences and matched odds ratios. R currently supports a wide variety of tools for the analysis of categorical data. However, many functions are spread across a variety of packages with differing syntax and poor compatibility with each another. prop_test() combines the functions binom.test(), prop.test() and BinomCI() into one output. prop_power() allows for power and sample size calculations for both balanced and unbalanced designs. riskdiff() is used for calculating risk differences and matched_or() is used for calculating matched odds ratios. For further information on methods used that are not documented in other packages see Nathan Mantel and William Haenszel (1959) <doi:10.1093/jnci/22.4.719> and Alan Agresti (2002) <ISBN:0-471-36093-7>. |
| License: | MIT + file LICENSE |
| Encoding: | UTF-8 |
| LazyData: | true |
| Imports: | epitools, DescTools, cli, magrittr, Hmisc, broom, rlang |
| RoxygenNote: | 6.1.1 |
| Suggests: | testthat, dplyr, forcats |
| NeedsCompilation: | no |
| Packaged: | 2019-06-14 13:52:19 UTC; niw4001 |
| Repository: | CRAN |
| Date/Publication: | 2019-06-14 14:10:03 UTC |
Pipe operator
Description
See magrittr::%>% for details.
Usage
lhs %>% rhs
Matched pairs odds ratio and confidence interval
Description
Create odds ratio and confidence interval from matched pairs data.
Usage
matched_or(df, ...)
Arguments
df |
a dataframe with binary variables x and y or a 2 x 2 frequency table/matrix. If a table or matrix, x and y must be NULL. Used to select method. |
... |
further arguments passed to or from other methods. |
Details
The matched pairs odds ratio and confidence interval is the equivalent of calculating a Cochran-Mantel-Haenszel odds ratio where each pair is treated as a stratum.
Value
a list with class "matched_or" with the following components:
tab |
2x2 table using for calculating risk difference |
or |
dataframe with columns corresponding to matched-pairs OR, lower bound, and upper bound of CI |
conf.level |
specified confidence level |
Examples
set.seed(1)
gene <- data.frame(pair = seq(1:35),
ulcer = rbinom(35, 1, .7),
healthy = rbinom(35, 1, .4))
matched_or(gene, ulcer, healthy)
Matched pairs odds ratio from a data frame
Description
Create odds ratio and confidence interval from matched pairs data.
Usage
## S3 method for class 'data.frame'
matched_or(df, x, y, weight = NULL, alpha = 0.05,
rev = c("neither", "rows", "columns", "both"), ...)
Arguments
df |
a dataframe with binary variables x and y. |
x |
binary vector, used as rows for frequency table and calculations. |
y |
binary vector, used as columns for frequency table and calculations. |
weight |
an optional vector of count weights. |
alpha |
level of significance for confidence interval. |
rev |
reverse order of cells. Options are "row", "columns", "both", and "neither" (default). |
... |
further arguments passed to or from other methods. |
Value
a list with class "matched_or" with the following components:
tab |
2x2 table using for calculating risk difference |
or |
dataframe with columns corresponding to matched-pairs OR, lower bound, and upper bound of CI |
conf.level |
specified confidence level |
Examples
gene <- data.frame(pair = seq(1:35),
ulcer = rbinom(35, 1, .7),
healthy = rbinom(35, 1, .4))
matched_or(gene, ulcer, healthy)
Matched pairs odds ratio from a table
Description
Create odds ratio and confidence interval from matched pairs data.
Usage
## S3 method for class 'table'
matched_or(df, alpha = 0.05, rev = c("neither", "rows",
"columns", "both"), ...)
Arguments
df |
a dataframe with binary variables x and y or a 2 x 2 frequency table/matrix. |
alpha |
level of significance for confidence interval. |
rev |
reverse order of cells. Options are "row", "columns", "both", and "neither" (default). |
... |
further arguments passed to or from other methods. |
Value
a list with class "matched_or" with the following components:
tab |
2x2 table using for calculating risk difference |
or |
dataframe with columns corresponding to matched-pairs OR, lower bound, and upper bound of CI |
conf.level |
specified confidence level |
Examples
gene <- data.frame(pair = seq(1:35),
ulcer = rbinom(35, 1, .7),
healthy = rbinom(35, 1, .4))
gene_tab <- xtabs(~ ulcer + healthy, data = gene)
gene_tab %>% matched_or()
Power and sample size for 2 proportions
Description
Calculate power and sample size for comparison of 2 proportions for both balanced and unbalanced designs.
Usage
prop_power(n, n1, n2, p1, p2, fraction = 0.5, alpha = 0.05,
power = NULL, alternative = c("two.sided", "one.sided"), odds.ratio,
percent.reduction, ...)
Arguments
n |
total sample size. |
n1 |
sample size in group 1. |
n2 |
sample size in group 2. |
p1 |
group 1 proportion. |
p2 |
group 2 proportion. |
fraction |
fraction of total observations that are in group 1. |
alpha |
significance level/type 1 error rate. |
power |
desired power, between 0 and 1. |
alternative |
alternative hypothesis, one- or two-sided test. |
odds.ratio |
odds ratio comparing p2 to p2. |
percent.reduction |
percent reduction of p1 to p2. |
... |
further arguments passed to or from other methods. |
Details
Power calculations are done using the methods described in 'stats::power.prop.test', 'Hmisc::bsamsize', and 'Hmisc::bpower'.
Value
a list with class "prop_power" containing the following components:
n |
the total sample size |
n1 |
the sample size in group 1 |
n2 |
the sample size in group 2 |
p1 |
the proportion in group 1 |
p2 |
the proportion in group 2 |
power |
calculated or desired power |
sig.level |
level of significance |
See Also
[stats::power.prop.test], [Hmisc::bsamsize], [Hmisc:bpower]
Examples
prop_power(n = 220, p1 = 0.35, p2 = 0.2)
prop_power(p1 = 0.35, p2 = 0.2, fraction = 2/3, power = 0.85)
prop_power(p1 = 0.35, n = 220, percent.reduction = 42.857)
prop_power(p1 = 0.35, n = 220, odds.ratio = 0.4642857)
Tests for equality of proportions
Description
Conduct 1-sample tests of proportions and tests for equality of k proportions.
Usage
prop_test(x, ...)
Arguments
x |
a vector of counts, a one-dimensional table with two entries, or a two-dimensional table with 2 columns. Used to select method. |
... |
further arguments passed to or from other methods. |
Details
Calculations are done using the methods described in 'stats::binom.test()' and 'stats::prop.test()'
Value
a list with class "prop_test" containing the following components:
x |
number of successes |
n |
number of trials |
p |
null proportion |
statistic |
the value of Pearson's chi-squared test statistic |
p_value |
p-value corresponding to chi-squared test statistic |
df |
degrees of freedom |
method |
the method used to calculate the confidence interval |
method_ci |
confidence interval calculated using specified method |
exact_ci |
exact confidence interval |
exact_p |
p-value from exact test |
See Also
[stats::binom.test()], [stats::prop.test()]
Examples
prop_test(7, 50, method = "wald", p = 0.2)
prop_test(7, 50, method = "wald", p = 0.2, exact = TRUE)
prop_test(c(23, 24), c(50, 55))
vietnam <- data.frame(
service = c(rep("yes", 2), rep("no", 2)),
sleep = c(rep(c("yes", "no"), 2)),
count = c(173, 160, 599, 851)
)
sleep <- xtabs(count ~ service + sleep, data = vietnam)
prop_test(sleep)
prop_test(vietnam, service, sleep, count)
Tests for equality of proportions
Description
Conduct 1-sample tests of proportions and tests for equality of k proportions.
Usage
## S3 method for class 'data.frame'
prop_test(x, pred, out, weight = NULL,
rev = c("neither", "rows", "columns", "both"), method = c("wald",
"wilson", "agresti-couli", "jeffreys", "modified wilson", "wilsoncc",
"modified jeffreys", "clopper-pearson", "arcsine", "logit", "witting",
"pratt"), alternative = c("two.sided", "less", "greater"),
conf.level = 0.95, correct = FALSE, exact = FALSE, ...)
Arguments
x |
a dataframe with categorical variable |
pred |
predictor/exposure, vector. |
out |
outcome, vector. |
weight |
an optional vector of count weights. |
rev |
reverse order of cells. Options are "row", "columns", "both", and "neither" (default). |
method |
a character string indicating method for calculating confidence interval, default is "wald". Options include, wald, wilson, agresti-couli, jeffreys, modified wilson, wilsoncc modified jeffreys, clopper-pearson, arcsine, logit, witting, and pratt. |
alternative |
character string specifying the alternative hypothesis. Possible options are "two.sided" (default), "greater", or "less". |
conf.level |
confidence level for confidence interval, default is 0.95. |
correct |
a logical indicating whether Yate's continuity correction should be applied. |
exact |
a logical indicating whether to output exact p-value, ignored if k-sample test. |
... |
further arguments passed to or from other methods. |
Value
a list with class "prop_test" containing the following components:
x |
number of successes |
n |
number of trials |
p |
null proportion |
statistic |
the value of Pearson's chi-squared test statistic |
p_value |
p-value corresponding to chi-squared test statistic |
df |
degrees of freedom |
method |
the method used to calculate the confidence interval |
method_ci |
confidence interval calculated using specified method |
exact_ci |
exact confidence interval |
exact_p |
p-value from exact test |
Examples
vietnam <- data.frame(
service = c(rep("yes", 2), rep("no", 2)),
sleep = c(rep(c("yes", "no"), 2)),
count = c(173, 160, 599, 851)
)
prop_test(vietnam, service, sleep, count)
Tests for equality of proportions
Description
Conduct 1-sample tests of proportions and tests for equality of k proportions.
Usage
## S3 method for class 'matrix'
prop_test(x, method = c("wald", "wilson",
"agresti-couli", "jeffreys", "modified wilson", "wilsoncc",
"modified jeffreys", "clopper-pearson", "arcsine", "logit", "witting",
"pratt"), alternative = c("two.sided", "less", "greater"),
conf.level = 0.95, correct = FALSE, exact = FALSE, ...)
Arguments
x |
a 2 x k matrix. |
method |
a character string indicating method for calculating confidence interval, default is "wald". Options include, wald, wilson, agresti-couli, jeffreys, modified wilson, wilsoncc modified jeffreys, clopper-pearson, arcsine, logit, witting, and pratt. |
alternative |
character string specifying the alternative hypothesis. Possible options are "two.sided" (default), "greater", or "less". |
conf.level |
confidence level for confidence interval, default is 0.95. |
correct |
a logical indicating whether Yate's continuity correction should be applied. |
exact |
a logical indicating whether to output exact p-value, ignored if k-sample test. |
... |
further arguments passed to or from other methods. |
Value
a list with class "prop_test" containing the following components:
x |
number of successes |
n |
number of trials |
p |
null proportion |
statistic |
the value of Pearson's chi-squared test statistic |
p_value |
p-value corresponding to chi-squared test statistic |
df |
degrees of freedom |
method |
the method used to calculate the confidence interval |
method_ci |
confidence interval calculated using specified method |
exact_ci |
exact confidence interval |
exact_p |
p-value from exact test |
Examples
matrix(c(23, 48, 76, 88), nrow = 2, ncol = 2) %>% prop_test()
Tests for equality of proportions
Description
Conduct 1-sample tests of proportions and tests for equality of k proportions.
Usage
## S3 method for class 'numeric'
prop_test(x, n, p = 0.5, method = c("wald", "wilson",
"agresti-couli", "jeffreys", "modified wilson", "wilsoncc",
"modified jeffreys", "clopper-pearson", "arcsine", "logit", "witting",
"pratt"), alternative = c("two.sided", "less", "greater"),
conf.level = 0.95, correct = FALSE, exact = FALSE, ...)
Arguments
x |
a vector of counts. |
n |
a vector of counts of trials |
p |
a probability for the null hypothesis when testing a single proportion; ignored if comparing multiple proportions. |
method |
a character string indicating method for calculating confidence interval, default is "wald". Options include, wald, wilson, agresti-couli, jeffreys, modified wilson, wilsoncc modified jeffreys, clopper-pearson, arcsine, logit, witting, and pratt. |
alternative |
character string specifying the alternative hypothesis. Possible options are "two.sided" (default), "greater", or "less". |
conf.level |
confidence level for confidence interval, default is 0.95. |
correct |
a logical indicating whether Yate's continuity correction should be applied. |
exact |
a logical indicating whether to output exact p-value, ignored if k-sample test. |
... |
further arguments passed to or from other methods. |
Value
a list with class "prop_test" containing the following components:
x |
number of successes |
n |
number of trials |
p |
null proportion |
statistic |
the value of Pearson's chi-squared test statistic |
p_value |
p-value corresponding to chi-squared test statistic |
df |
degrees of freedom |
method |
the method used to calculate the confidence interval |
method_ci |
confidence interval calculated using specified method |
exact_ci |
exact confidence interval |
exact_p |
p-value from exact test |
Examples
prop_test(7, 50, method = "wald", p = 0.2)
prop_test(7, 50, method = "wald", p = 0.2, exact = TRUE)
Tests for equality of proportions
Description
Conduct 1-sample tests of proportions and tests for equality of k proportions.
Usage
## S3 method for class 'table'
prop_test(x, method = c("wald", "wilson",
"agresti-couli", "jeffreys", "modified wilson", "wilsoncc",
"modified jeffreys", "clopper-pearson", "arcsine", "logit", "witting",
"pratt"), alternative = c("two.sided", "less", "greater"),
conf.level = 0.95, correct = FALSE, exact = FALSE, ...)
Arguments
x |
a 2 x k table. |
method |
a character string indicating method for calculating confidence interval, default is "wald". Options include, wald, wilson, agresti-couli, jeffreys, modified wilson, wilsoncc modified jeffreys, clopper-pearson, arcsine, logit, witting, and pratt. |
alternative |
character string specifying the alternative hypothesis. Possible options are "two.sided" (default), "greater", or "less". |
conf.level |
confidence level for confidence interval, default is 0.95. |
correct |
a logical indicating whether Yate's continuity correction should be applied. |
exact |
a logical indicating whether to output exact p-value, ignored if k-sample test. |
... |
further arguments passed to or from other methods. |
Value
a list with class "prop_test" containing the following components:
x |
number of successes |
n |
number of trials |
p |
null proportion |
statistic |
the value of Pearson's chi-squared test statistic |
p_value |
p-value corresponding to chi-squared test statistic |
df |
degrees of freedom |
method |
the method used to calculate the confidence interval |
method_ci |
confidence interval calculated using specified method |
exact_ci |
exact confidence interval |
exact_p |
p-value from exact test |
Examples
vietnam <- data.frame(
service = c(rep("yes", 2), rep("no", 2), rep("maybe", 2)),
sleep = rep(c("yes", "no"), 3),
count = c(173, 160, 599, 851, 400, 212)
)
xtabs(count ~ service + sleep, data = vietnam) %>% prop_test()
Risk difference
Description
Calculate risk difference and 95 percent confidence interval using Wald method.
Usage
riskdiff(df, ...)
Arguments
df |
a dataframe with binary variables x and y or a 2 x 2 frequency table/matrix. If a table or matrix, x and y must be NULL. Used to select method. |
... |
further arguments passed to or from other methods. |
Value
a list with class "rdiff" containing the following components:
rd |
risk difference |
conf.level |
specified confidence level |
ci |
calculated confidence interval |
p1 |
proportion one |
p2 |
proportion two |
tab |
2x2 table using for calculating risk difference |
Examples
trial <- data.frame(
disease = c(rep("yes", 2), rep("no", 2)),
treatment = c(rep(c("estrogen", "placebo"), 2)),
count = c(751, 623, 7755, 7479))
riskdiff(trial, treatment, disease, count, rev = "columns")
Risk difference
Description
Calculate risk difference and 95 percent confidence interval using Wald method.
Usage
## S3 method for class 'data.frame'
riskdiff(df, x = NULL, y = NULL, weight = NULL,
conf.level = 0.95, rev = c("neither", "rows", "columns", "both"),
...)
Arguments
df |
a dataframe with binary variables x and y. |
x |
binary predictor/exposure, vector. |
y |
binary outcome, vector. |
weight |
an optional vector of count weights. |
conf.level |
confidence level for confidence interval, default is 0.95. |
rev |
reverse order of cells. Options are "row", "columns", "both", and "neither" (default). |
... |
further arguments passed to or from other methods. |
Value
a list with class "rdiff" containing the following components:
rd |
risk difference |
conf.level |
specified confidence level |
ci |
calculated confidence interval |
p1 |
proportion one |
p2 |
proportion two |
tab |
2x2 table using for calculating risk difference |
Examples
trial <- data.frame(
disease = c(rep("yes", 2), rep("no", 2)),
treatment = c(rep(c("estrogen", "placebo"), 2)),
count = c(751, 623, 7755, 7479))
riskdiff(trial, treatment, disease, count, rev = "columns")
Risk difference
Description
Calculate risk difference and 95 percent confidence interval using Wald method.
Usage
## S3 method for class 'matrix'
riskdiff(df, conf.level = 0.95, dnn = NULL,
rev = c("neither", "rows", "columns", "both"), ...)
Arguments
df |
a 2 x 2 frequency matrix. |
conf.level |
confidence level for confidence interval, default is 0.95. |
dnn |
optional character vector of dimension names. |
rev |
reverse order of cells. Options are "row", "columns", "both", and "neither" (default). |
... |
further arguments passed to or from other methods. |
Value
a list with class "rdiff" containing the following components:
rd |
risk difference |
conf.level |
specified confidence level |
ci |
calculated confidence interval |
p1 |
proportion one |
p2 |
proportion two |
tab |
2x2 table using for calculating risk difference |
Examples
matrix(c(12, 45, 69, 15), nrow = 2, ncol = 2) %>%
riskdiff(dnn = c("New Drug", "Adverse Outcome"))
Risk difference
Description
Calculate risk difference and 95 percent confidence interval using Wald method.
Usage
## S3 method for class 'table'
riskdiff(df, conf.level = 0.95, rev = c("neither",
"rows", "columns", "both"), ...)
Arguments
df |
a 2 x 2 frequency table. |
conf.level |
confidence level for confidence interval, default is 0.95. |
rev |
reverse order of cells. Options are "row", "columns", "both", and "neither" (default). |
... |
further arguments passed to or from other methods. |
Value
a list with class "rdiff" containing the following components:
rd |
risk difference |
conf.level |
specified confidence level |
ci |
calculated confidence interval |
p1 |
proportion one |
p2 |
proportion two |
tab |
2x2 table using for calculating risk difference |
Examples
trial <- data.frame(
disease = c(rep("yes", 2), rep("no", 2)),
treatment = c(rep(c("estrogen", "placebo"), 2)),
count = c(751, 623, 7755, 7479))
xtabs(count ~ treatment + disease, data = trial) %>% riskdiff()
Create 2 x k frequency tables
Description
Helper function for creating 2 x k frequency tables.
Usage
tavolo(df, ...)
Arguments
df |
a dataframe with binary variable y and categorical variable x or a 2 x k frequency table/matrix. If a table or matrix, x and y must be NULL. Used to select method. |
... |
further arguments passed to or from other methods. |
Value
tab |
2 x k frequency table |
Examples
trial <- data.frame(disease = c(rep("yes", 2), rep("no", 2)),
treatment = c(rep(c("estrogen", "placebo"), 2)),
count = c(751, 623, 7755, 7479))
tavolo(trial, treatment, disease, count)
Create 2 x k frequency tables
Description
Helper function for creating 2 x k frequency tables.
Usage
## S3 method for class 'data.frame'
tavolo(df, x, y, weight = NULL, rev = c("neither",
"rows", "columns", "both"), ...)
Arguments
df |
a dataframe with binary variable y and categorical variable x. |
x |
categorical predictor/exposure, vector. |
y |
binary outcome, vector. |
weight |
an optional vector of count weights. |
rev |
character string indicating whether to switch row or column order, possible options are "neither", "rows", "columns", or "both". The default is "neither". |
... |
further arguments passed to or from other methods. |
Value
tab |
2 x k frequency table |
Examples
trial <- data.frame(disease = c(rep("yes", 2), rep("no", 2)),
treatment = c(rep(c("estrogen", "placebo"), 2)),
count = c(751, 623, 7755, 7479))
tavolo(trial, treatment, disease, count)
Create 2 x k frequency tables
Description
Helper function for creating 2 x k frequency tables.
Usage
## S3 method for class 'matrix'
tavolo(df, dnn = NULL, rev = c("neither", "rows",
"columns", "both"), ...)
Arguments
df |
a 2 x k frequency matrix. |
dnn |
optional character vector of dimension names. |
rev |
character string indicating whether to switch row or column order, possible options are "neither", "rows", "columns", or "both". The default is "neither". |
... |
further arguments passed to or from other methods. |
Value
tab |
2 x k frequency table |
Examples
tavolo(matrix(c(23, 45, 67, 12), nrow = 2, ncol = 2), rev = "both")
Create 2 x k frequency tables
Description
Helper function for creating 2 x k frequency tables.
Usage
## S3 method for class 'table'
tavolo(df, rev = c("neither", "rows", "columns", "both"),
...)
Arguments
df |
a 2 x k frequency table. |
rev |
character string indicating whether to switch row or column order, possible options are "neither", "rows", "columns", or "both". The default is "neither". |
... |
further arguments passed to or from other methods. |
Value
tab |
2 x k frequency table |
Examples
trial <- data.frame(disease = c(rep("yes", 3), rep("no", 3)),
treatment = rep(c("estrogen", "placebo", "other"), 2),
count = c(751, 623, 7755, 7479, 9000, 456))
xtabs(count ~ treatment + disease, data = trial) %>% tavolo(rev = "columns")