scratch_pad.R

In PiotrTymoszuk/ExDA: Toolset for Explorative Data Analysis

# Tests of the package

# required tools ------

  library(tidyverse)
  library(generics)
  library(exda)

  explore <- exda::explore

# behavior of the EDA objects

  test1 <- eda(c(rep('a', 10), rep('b', 5), rep('c', 3)))
  test2 <- eda(c(rep('a', 5), rep('b', 5), rep('c', 8)))
  test5 <- eda(c(rep('a', 1), rep('b', 2), rep('d', 8), NA))

  summary(test5, pub_styled = TRUE)

  summary(distribution(test1, test2))

  summary(correlation(test1, test2, type = 'kappa'), pub_styled = TRUE)

  covariance(test1, test2)

  #variance(test1, test2)

  summary(test(test2, test5, type = 'chisq_test'), pub_styled = TRUE)

  test3 <- eda(sort(c(NA, rnorm(100))))
  test4 <- eda(sort(c(rt(100, df = 10), NA)))
  test6 <- eda(sort(c(rchisq(100, df = 10), NA)))

  summary(test3)

  summary(eda(c(NA, rnorm(100))), pub_styled = TRUE)


  summary(normality(test3), pub_styled = TRUE)
  normality(test4)

  summary(distribution(test3, test4), pub_styled = TRUE)

  summary(correlation(test3, test4, type = 'kendall'), pub_styled = TRUE)

  #as_tibble(test3)

  variance(test3, test4)

  covariance(test3, test4, 'kendall')

  summary(test(test3, test4, test6, type = 'friedman_test', ci = FALSE),
          pub_styled = TRUE,
          signif_digits = 2,
          simplify_p = TRUE)

  plot(test6, type = 'hist')

# Descriptive stats and hypothesis testing -------

  explore(mtcars, what = 'table', signif_digits = 4)

  explore(data = mtcars, split_factor = 'am', what = 'plots', type = 'violin')

  explore(data = mtcars, split_factor = 'am', what = 'table')

  compare_variables(mtcars[1:10, ], mtcars[11:20, ], mtcars[21:32, ],
                    variables = c('mpg', 'disp', 'wt', 'qsec'),
                    what = 'test',
                    types = 'anova',
                    pub_styled = TRUE,
                    adj_method = 'holm')

  compare_variables(mtcars[1:10, ], mtcars[11:20, ], mtcars[21:32, ],
                    variables = 'mpg',
                    what = 'correlation',
                    types = 'pearson')

  new_cars <- mtcars %>%
    as_tibble %>%
    mutate(vs = factor(vs),
           am = factor(am),
           gear = factor(gear),
           grouping = sample(c('a', 'b'), 32, replace = TRUE),
           corr_grouping = rep(c('c', 'd'), 16))


  exda:::compare(new_cars, variable = 'gear', split_factor = 'vs', type = 'kruskal_test')


  compare_variables(new_cars[1:10, ], new_cars[11:20, ], new_cars[21:30, ],
                    variables = c('mpg', 'vs', 'gear', 'am'),
                    what = 'correlation',
                    types = c('pearson', 'kendall', 'kendall', 'kendall'))

  compare_variables(new_cars[1:10, ], new_cars[11:20, ], new_cars[21:30, ],
                    variables = c('mpg', 'vs', 'gear', 'am', 'qsec', 'hp', 'carb'),
                    what = 'test',
                    types = c('mann_whitney_test', 'chisq_test', 'kruskal_test',
                              'chisq_test', 'rm_anova', 'kruskal_test', 'friedman_test'),
                    pub_styled = TRUE,
                    adj_method = 'holm',
                    .parallel = TRUE,
                    .paropts = furrr::furrr_options(seed = TRUE, globals = c('new_cars')))

  compare_variables(new_cars,
                    variables = c('mpg', 'vs', 'gear', 'am', 'qsec', 'hp', 'carb'),
                    split_factor = 'gear',
                    types = c('mann_whitney_test', 'chisq_test', 'chisq_test',
                              'chisq_test', 'anova', 'kruskal_test', 'anova'),
                    pub_styled = TRUE,
                    adj_method = 'holm',
                    ci = FALSE)

# Plotting ---------

  ## via multiplot

  multiplot(test1, test2, test5,
            type = 'bar',
            eda_names = c('t1', 't2', 't3'),
            scale = 'percent')

  multiplot(test3, test4, test6,
            type = 'paired',
            eda_names = c('t1', 't2', 't3'),
            bins = 10,
            facet_hist = 'vertical')

  ## plotting via plot_variable

  plot_variable(new_cars[1:10, ], new_cars[11:20, ], new_cars[21:30, ],
                variable = 'mpg',
                type = 'correlation')

  plot_variable(new_cars[1:10, ], new_cars[11:20, ], new_cars[21:30, ],
                variable = 'mpg',
                type = 'violin')

  plot_variable(new_cars, variable = 'mpg',
                split_factor = 'gear', type = 'violin')

  plot_variable(new_cars, variable = 'mpg',
                split_factor = 'corr_grouping', type = 'correlation')

# Correlations --------

  correlate_variables(new_cars[1:16, ],
                      new_cars[17:32, ],
                      variables = c('mpg', 'hp', 'qsec'),
                      what = 'correlation',
                      type = 'spearman')

  correlate_variables(new_cars[1:16, ],
                      new_cars[17:32, ],
                      variables = c('vs', 'am', 'qsec'),
                      what = 'correlation',
                      type = 'kappa')

  plot_correlation(new_cars,
                   variables = c('vs', 'am'),
                   type = 'bar',
                   scale = 'percent')

  multiplot(test1, test2, test5,
            type = 'bubble',
            eda_names = c('t1', 't2', 't3'),
            scale = 'none')

  #multiplot(test1, eda(c(rep('e', 6), rep('a', 6), rep('b', 5), NA)),
     #       type = 'bubble',
      #      eda_names = c('t1', 't5'),
      #      scale = 'percent')

  plot_correlation(new_cars,
                   variables = c('vs', 'am'),
                   type = 'bubble',
                   scale = 'none')

# Effect size --------

  eff_size(test1, test2, type = 'freeman_theta',
           ci = FALSE, boot_method = 'bca')

  eff_size(test1, test5,
           type = 'cramer_v', ci = TRUE)

  eff_size(test1, test2,
           type = 'cramer_v', ci = TRUE)

  eff_size(test3, test4,
           type = 'paired_cohen_d')

  eff_size(test3, test6,
           type = 'paired_wilcoxon_r',
           ci = TRUE,
           boot_method = 'normality')

  eff_size(test1, test2, test5,
           type = 'cohen_kappa')

  eff_size(test3, test4, test6,
           type = 'kendall_w')

  compare_variables(new_cars,
                    what = 'eff_size',
                    variables = c('mpg', 'vs', 'gear', 'am',
                                  'qsec', 'hp', 'carb'),
                    split_factor = 'gear',
                    types = c('wilcoxon_r', 'cramer_v', 'cramer_v',
                              'cramer_v', 'etasq', 'kruskal_etasq', 'etasq'),
                    pub_styled = TRUE,
                    adj_method = 'holm',
                    ci = FALSE)

  test_eff <- compare_variables(new_cars[1:16, ],
                                new_cars[17:32, ],
                                variables = c('vs', 'am', 'qsec'),
                                what = 'eff_size',
                                types = c('cohen_kappa', 'cramer_v',
                                          'paired_cohen_d'))

# Additional plotting options: stack plot -------

  plot(test_eff, show_labels = 'signif')

  multiplot(test1, test2, test5,
            type = 'stack',
            eda_names = c('t1', 't2', 't3'),
            scale = 'percent')

  plot_variable(new_cars[1:10, ], new_cars[11:20, ], new_cars[21:30, ],
                variable = 'gear',
                type = 'bar')

  plot_variable(new_cars[1:10, ], new_cars[11:20, ], new_cars[21:30, ],
                variable = 'gear',
                type = 'stack')

  plot_variable(new_cars,
                variable = 'gear',
                split_factor = 'grouping',
                type = 'stack',
                scale = 'fraction')

# Plot panels -------

  draw_quantile_elli(mtcars,
                     quantiles = c(0.025, 0.975),
                     variables = c('disp', 'wt', 'hp'))

  draw_quantile_elli(dplyr::mutate(mtcars, vs = factor(vs, c('1', '0'))),
                     quantiles = c(0.025, 0.975),
                     variables = c('disp', 'wt', 'hp'),
                     split_factor = 'vs')

  draw_violin_panel(mtcars,
                    variables = c('disp', 'hp'))

  draw_violin_panel(dplyr::mutate(mtcars, vs = factor(vs, c('0', '1'))),
                    variables = c('disp', 'hp'),
                    split_factor = 'vs',
                    distr_geom = 'box',
                    alpha = 1)

  draw_stat_panel(dplyr::mutate(mtcars,
                                gear_fct = factor(gear, c(4, 5, 3)),
                                vs = factor(vs, c('1', '0'))),
                  variables = c('disp', 'hp', 'wt', 'mpg', 'qsec', 'carb'),
                  stat = 'mean',
                  split_factor = 'vs',
                  err_stat = '2se',
                  form = 'line')

  plot_variable(dplyr::mutate(mtcars,
                              gear_fct = factor(gear, c(4, 5, 3)),
                              vs = factor(vs, c('1', '0'))),
                split_factor = 'vs',
                variable = 'gear_fct',
                type = 'stack',
                scale = 'percent',
                txt_color = 'orangered3')

  plot_variable(dplyr::mutate(mtcars,
                              gear_fct = factor(gear, c(4, 5, 3)),
                              vs = factor(vs, c('1', '0'))),
                split_factor = 'vs',
                variable = 'mpg',
                type = 'violin',
                scale = 'percent',
                txt_color = 'orangered3')

  draw_freq_panel(dplyr::mutate(mtcars,
                                gear_fct = factor(gear, c(4, 5, 3)),
                                vs = factor(vs, c('1', '0')),
                                am = factor(am, c('1', '0'))),
                  #split_factor = 'cyl',
                  variables = c('vs', 'am'),
                  show_labels = TRUE)

# Testing the 'x_n_labs' option: n numbers in the X axis ---------

  plot_variable(dplyr::mutate(mtcars,
                              gear_fct = factor(gear, c(4, 5, 3)),
                              vs = factor(vs, c('1', '0'))),
                split_factor = 'vs',
                variable = 'mpg',
                type = 'violin',
                scale = 'percent',
                txt_color = 'orangered3',
                x_n_labs = TRUE)

  plot_variable(new_cars,
                variable = 'gear',
                split_factor = 'grouping',
                type = 'stack',
                scale = 'fraction',
                x_n_labs = TRUE)

  plot_variable(new_cars,
                variable = 'gear',
                split_factor = 'grouping',
                type = 'bar',
                scale = 'fraction',
                x_n_labs = TRUE)

  plot_variable(new_cars,
                variable = 'wt',
                split_factor = 'corr_grouping',
                type = 'paired',
                x_n_labs = TRUE)

# END ------