R/descriptives.R

In samuelkaminsky/SKTools: SK's Tools

#' Dataframe Descriptives and Frequencies
#' @param df Dataframe with data
#' @param frequencies Logical to indicate whether or not to calculate frequencies for each variable as a list column
#' @return Dataframe of descriptives and frequencies for each variable and value
#' @description Calculates descriptives and frequencies of every Column in a dataframe. Heavily inspired by Ujjwal Karn's XDA package.
#' @export

descriptives <-
  function(df, frequencies = FALSE) {
    df.func <-
      df %>%
      tibble::set_tidy_names(syntactic = TRUE, quiet = TRUE) %>%
      dplyr::mutate_if(is.factor, as.character) %>%
      sjlabelled::remove_all_labels()

    if (isTRUE(frequencies)) {
      freqs <-
        df.func %>% 
        names() %>% 
        purrr::set_names() %>% 
        purrr::map_dfr(function(x) {
          df.func %>%
            dplyr::count(!!x) %>%
            purrr::set_names(c("value", "n")) %>%
            dplyr::mutate(value = as.character(.data$value))
        }, .id = "var") %>%
        dplyr::group_by(.data$var) %>%
        tidyr::nest(.key = "frequencies") %>%
        dplyr::mutate(frequencies = .data$frequencies %>% purrr::set_names(.data$var)) %>%
        dplyr::arrange(.data$var)
    }

    missing <-
      df.func %>%
      dplyr::mutate_all(list(as.character)) %>%
      tidyr::gather("var", "value") %>%
      dplyr::group_by(.data$var) %>%
      dplyr::summarize(
        n = sum(!is.na(.data$value)),
        n_missing = sum(is.na(.data$value)),
        perc_missing = .data$n_missing / dplyr::n()
      )

    class <-
      df.func %>%
      purrr::map(class) %>%
      purrr::map_dfr(stringr::str_c, collapse = ", ") %>%
      tidyr::gather("var", "class")

    df.func <-
      df.func %>%
      dplyr::select_if(is.numeric) %>%
      tidyr::gather("var", "value") %>%
      dplyr::group_by(.data$var) %>%
      dplyr::summarize(
        mean = mean(.data$value, na.rm = TRUE),
        median = stats::median(.data$value, na.rm = TRUE),
        max = max(.data$value, na.rm = TRUE),
        min = min(.data$value, na.rm = TRUE),
        sd = stats::sd(.data$value, na.rm = TRUE),
        range = .data$max - .data$min,
        iqr = stats::IQR(.data$value, na.rm = TRUE),
        skewness = moments::skewness(.data$value, na.rm = TRUE),
        kurtosis = moments::kurtosis(.data$value, na.rm = TRUE),
        n_unique = length(unique(stats::na.omit(.data$value))),
        `1%` = stats::quantile(.data$value, .01, na.rm = TRUE),
        `25%` = stats::quantile(.data$value, .25, na.rm = TRUE),
        `50%` = stats::quantile(.data$value, .5, na.rm = TRUE),
        `75%` = stats::quantile(.data$value, .75, na.rm = TRUE),
        `99%` = stats::quantile(.data$value, .99, na.rm = TRUE)
      ) %>%
      dplyr::full_join(missing, by = "var", na_matches = "never") %>%
      dplyr::full_join(class, by = "var", na_matches = "never")

    if (isTRUE(frequencies)) {
      df.func <-
        df.func %>%
        dplyr::full_join(freqs, by = "var", na_matches = "never") %>%
        dplyr::mutate(frequencies = .data$frequencies %>% purrr::set_names(.data$var))
    }
    if (isTRUE(frequencies)) {
      df.func %>%
        dplyr::select(
          .data$var,
          .data$class,
          .data$mean:.data$n_unique,
          .data$n,
          .data$n_missing,
          .data$perc_missing,
          .data$`1%`:.data$`99%`,
          .data$frequencies
        )
    } else {
      df.func %>%
        dplyr::select(
          .data$var,
          .data$class,
          .data$mean:.data$n_unique,
          .data$n,
          .data$n_missing,
          .data$perc_missing,
          .data$`1%`:.data$`99%`
        )
    }
  }