Syntactic Sugar for Object Construction

Construct test objects in a unified way.

Usage

pmt(key, ...)

pmts(
  which = c("all", "onesample", "twosample", "distribution", "association", "paired",
    "ksample", "multcomp", "rcbd", "table")
)

define_pmt(
  method = c("twosample", "distribution", "association", "paired", "ksample", "rcbd",
    "table"),
  statistic,
  rejection = c("<>", "<", ">"),
  scoring = "none",
  n_permu = 10000,
  name = "User-Defined Permutation Test",
  alternative = NULL,
  quickr = FALSE,
  depends = character(),
  plugins = character(),
  includes = character()
)

Arguments

key: a character string specifying the test. Check pmts() for valid keys.
...: extra parameters passed to the constructor.
which: a character string specifying the desired tests.
method: a character string specifying the permutation scheme.
statistic: definition of the test statistic. See details.
rejection: a character string specifying the rejection region relative to the test statistic.
scoring: one of: - a character string in c("none", "rank", "vw", "expon") specifying the scoring system - a function that takes a numeric vector and returns an equal-length score vector
n_permu: an integer indicating number of permutations for the permutation distribution. If set to 0, all permutations will be used.
name, alternative: character strings specifying the name of the test and the alternative hypothesis, used for printing purposes only.
quickr: a logical indicating whether to use quickr::quick() to accelerate statistic. See details.
depends, plugins, includes: passed to Rcpp::cppFunction().

Value

a test object corresponding to the specified key.

a data frame containing keys and corresponding tests implemented in this package.

a test object based on the specified statistic.

Details

The test statistic can be defined using either R or Rcpp, with the statistic parameter specified as:

R: a closure returning one of
- a double (the test statistic).
- a closure returning a double.
Rcpp: a character string defining a captureless lambda (since C++11) returning another lambda that captures by value, accepts parameters of the same type, and returns a double.

This design aims to pre-calculate potential constants that remain invariant during permutation.

When using Rcpp, the parameters for different method are listed as follows. Note that the names can be customized, and the types can be replaced with auto (thanks to the support for generic lambdas in C++14). See examples.

`method`	Parameter 1	Parameter 2
`"twosample"`	`const NumericVector& sample_1`	`const NumericVector& sample_2`
`"distribution"`	`const NumericVector& cumulative_prob_1`	`const NumericVector& cumulative_prob_2`
`"association"`	`const NumericVector& sample_1`	`const NumericVector& sample_2`
`"paired"`	`const NumericVector& sample_1`	`const NumericVector& sample_2`
`"ksample"`	`const NumericVector& combined_sample`	`const IntegerVector& one_based_group_index`
`"rcbd"`	`const NumericMatrix& block_as_column_data`
`"table"`	`const IntegerMatrix& contingency_table`

When using R, statistic and the parameters should be the R equivalents of the above. If no constants exist during permutation, statistic may simply be an R closure returning a double.

If quickr = TRUE and statistic returns a double, it will be compiled to Fortran via quickr::quick() with base::declare() calls for all arguments inserted automatically. Otherwise, statistic will be compiled using compiler::cmpfun().

Note

To improve performance when calling R closures from C++, this package repeatedly evaluates the closure's body in an environment whose enclosing environment is the closure's own, with its formal arguments pre-assigned to the data. This imposes the following restrictions on the closure returning the test statistic when statistic is written in R:

Do not re-assign its formal arguments or any pre-computed symbols in its environment.
Do not use default arguments or variadic arguments.

It's also worth noting that the data is permuted in-place. Therefore, modifications to the data within statistic may lead to incorrect results. It is recommended to avoid modifying the data when using R and pass const references as in the table above when using Rcpp.

Examples

pmt("twosample.wilcoxon")
#> <Wilcoxon>
#>   Inherits from: <TwoSampleLocationTest>
#>   Public:
#>     alternative: active binding
#>     conf_int: active binding
#>     conf_level: active binding
#>     correct: active binding
#>     data: active binding
#>     estimate: active binding
#>     initialize: function (type = c("permu", "asymp"), alternative = c("two_sided", 
#>     method: active binding
#>     n_permu: active binding
#>     null_value: active binding
#>     p_value: active binding
#>     plot: function (style = c("graphics", "ggplot2"), ...) 
#>     print: function () 
#>     scoring: active binding
#>     statistic: active binding
#>     test: function (...) 
#>     type: active binding
#>   Private:
#>     .alternative: two_sided
#>     .autoplot: function (...) 
#>     .calculate: function () 
#>     .calculate_extra: function () 
#>     .calculate_n_permu: function () 
#>     .calculate_p: function () 
#>     .calculate_p_permu: function () 
#>     .calculate_score: function () 
#>     .calculate_side: function () 
#>     .calculate_statistic: function () 
#>     .compile: function () 
#>     .conf_int: NULL
#>     .conf_level: 0.95
#>     .correct: TRUE
#>     .data: NULL
#>     .define: function () 
#>     .estimate: NULL
#>     .link: +
#>     .method: default
#>     .n_permu: 10000
#>     .name: Two-Sample Wilcoxon Test
#>     .null_value: 0
#>     .on_alternative_change: function () 
#>     .on_conf_level_change: function () 
#>     .on_method_change: function () 
#>     .on_n_permu_change: function () 
#>     .on_null_value_change: function () 
#>     .on_scoring_change: function () 
#>     .on_type_change: function () 
#>     .p_value: NULL
#>     .param_name: location shift
#>     .plot: function (...) 
#>     .preprocess: function () 
#>     .print: function () 
#>     .raw_data: NULL
#>     .scoring: rank
#>     .side: NULL
#>     .statistic: NULL
#>     .statistic_func: NULL
#>     .type: permu

pmts("ksample")
#>              key              class                         test
#> 1 ksample.oneway             OneWay One-Way Test for Equal Means
#> 2     ksample.kw      KruskalWallis          Kruskal-Wallis Test
#> 3     ksample.jt JonckheereTerpstra     Jonckheere-Terpstra Test

x <- rnorm(5)
y <- rnorm(5, 1)

t <- define_pmt(
    method = "twosample", rejection = "<",
    scoring = base::rank, # equivalent to "rank"
    statistic = function(x, y) sum(x)
)$test(x, y)$print()
#> 
#>  	 User-Defined Permutation Test 
#> 
#> scoring: custom    type: permu(10000)    method: twosample
#> statistic = 28, p-value = 0.5785 (± 0.009678288 at 95% confidence)

t$scoring <- function(x) qnorm(rank(x) / (length(x) + 1)) # equivalent to "vw"
t$print()
#> 
#>  	 User-Defined Permutation Test 
#> 
#> scoring: custom    type: permu(10000)    method: twosample
#> statistic = 0.1141853, p-value = 0.5487 (± 0.009753225 at 95% confidence)

t$n_permu <- 0
t$print()
#> 
#>  	 User-Defined Permutation Test 
#> 
#> scoring: custom    type: permu(252)    method: twosample
#> statistic = 0.1141853, p-value = 0.5436508

# \donttest{
r <- define_pmt(
    method = "twosample", n_permu = 1e5,
    statistic = function(x, y) {
        m <- length(x)
        n <- length(y)
        function(x, y) sum(x) / m - sum(y) / n
    }
)

quickr <- define_pmt(
    method = "twosample", n_permu = 1e5, quickr = TRUE,
    statistic = function(x, y) sum(x) / length(x) - sum(y) / length(y)
)

rcpp <- define_pmt(
    method = "twosample", n_permu = 1e5,
    statistic = "[](const auto& x, const auto& y) {
        auto m = x.length();
        auto n = y.length();
        return [=](const auto& x, const auto& y) {
            return sum(x) / m - sum(y) / n;
        };
    }"
)

# equivalent
# rcpp <- define_pmt(
#     method = "twosample", n_permu = 1e5,
#     statistic = "[](const NumericVector& x, const NumericVector& y) {
#         R_xlen_t m = x.length();
#         R_xlen_t n = y.length();
#         return [m, n](const NumericVector& x, const NumericVector& y) -> double {
#             return sum(x) / m - sum(y) / n;
#         };
#     }"
# )

set.seed(0)
r$test(x, y)$print()
#> 
#>  	 User-Defined Permutation Test 
#> 
#> scoring: none    type: permu(1e+05)    method: twosample
#> statistic = 0.002255881, p-value = 1 (± 0 at 95% confidence)
set.seed(0)
quickr$test(x, y)$print()
#> Loading required namespace: quickr
#> 
#>  	 User-Defined Permutation Test 
#> 
#> scoring: none    type: permu(1e+05)    method: twosample
#> statistic = 0.002255881, p-value = 1 (± 0 at 95% confidence)
set.seed(0)
rcpp$test(x, y)$print()
#> 
#>  	 User-Defined Permutation Test 
#> 
#> scoring: none    type: permu(1e+05)    method: twosample
#> statistic = 0.002255881, p-value = 1 (± 0 at 95% confidence)

options(LearnNonparam.pmt_progress = FALSE)
system.time(r$test(x, y))
#>    user  system elapsed 
#>   0.114   0.000   0.114 
system.time(quickr$test(x, y))
#>    user  system elapsed 
#>   0.030   0.000   0.031 
system.time(rcpp$test(x, y))
#>    user  system elapsed 
#>   0.007   0.000   0.007 
# }