epi_stats_numeric: Get summary descriptive statistics for numeric columns.

In AntonioJBT/episcout: Quickly Clean, Explore and Visualise Large Epidemiological Datasets

Get summary descriptive statistics for numeric columns.

Description

epi_stats_numeric() generates several descriptive summary statistics for numeric variables.

Usage

epi_stats_numeric(num_vec = NULL, na.rm = TRUE, coef = 1.5, ...)

Arguments

num_vec	Numeric vector to test.
na.rm	Remove NA values, default is TRUE
coef	Coefficient for outlier detection, default is 1.5
...	is passed to skewness() and kurtosis()

Value

Data.frame with results from tests performed

Note

Normality is tested with Shapiro-Wilk (small values indicate non-normality). Testing normality can be contentious, likely uninformative and with Shapiro-Wilk can only be done for sample size between 3-5000. The package e1071 is used for skewness and kurtosis. For skewness: negative/longer left tail, positive/longer right tail,1 usually means non-normality. For kurtosis consider lower values,broader shape and longer tails (platy ~<3), normal (meso ~3)and slender/no tails (lepto ~<3). Outliers are detected with the Tukeymethod (above and below 1.5 * IQR). na.rm is TRUE by default for all tests.

Author(s)

Antonio J Berlanga-Taylor <\url{https://github.com/AntonioJBT/episcout}>

See Also

epi_stats_summary, epi_stats_format, epi_stats_tidy, skewness, kurtosis.

Examples

#####
# Load libraries needed:
library(episcout)
library(dplyr)
library(purrr)
library(e1071)
library(tibble)
library(tidyr)
#####

#####
# Generate a data frame:
n <- 1000
df <- data.frame(var_id = rep(1:(n / 2), each = 2),
                 var_to_rep = rep(c("Pre", "Post"), n / 2),
                 x = rnorm(n),
                 y = rbinom(n, 1, 0.50),
                 z = rpois(n, 2)
                )

# Explore first and last rows for first columns:
epi_head_and_tail(df)

# Add character/factor columns:
col_chr <- data.frame('chr1' = rep(c('A', 'B'), length.out = 1000),
                      'chr2' = rep(c('C', 'D'), length.out = 1000)
                      )
dim(col_chr)
df_cont_chr <- tibble::as.tibble(cbind(df, col_chr))
epi_head_and_tail(df_cont_chr)
epi_head_and_tail(df_cont_chr, last_cols = TRUE)

# Check variable types are what you expect:
epi_clean_count_classes(df_cont_chr)
str(df_cont_chr)
dim(df_cont_chr)
# var_id, y and z can be treated as factors or characters.
summary(as.factor(df_cont_chr$y))
summary(as.factor(df_cont_chr$z))
# Here we'll only transform y though:
df_cont_chr$y <- as.factor(df_cont_chr$y)
epi_clean_count_classes(df_cont_chr)
str(df_cont_chr)

# Designate some values as codes to be counted separately:
codes <- c('Pre', 'A', 'C', '1', '3')
#####

#####
# Count when codes are present, pass these as character or factor, specify
#  action is to count codes only:
stat_sum1 <- epi_stats_summary(df = df_cont_chr,
                               codes = codes,
                               class_type = 'chr_fct',
                               action = 'codes_only'
                               )
class(stat_sum1)
stat_sum1
#####

#####
# Add total for percentage calculation and order column to tidy up results:
perc_n <- nrow(df_cont_chr)
order_by <- 'percent'
stat_sum_tidy <- epi_stats_tidy(sum_df = stat_sum1,
                                order_by = order_by,
                                perc_n = perc_n
                                )
stat_sum_tidy
# Format them if needed:
epi_stats_format(stat_sum_tidy, digits = 0)
epi_stats_format(stat_sum_tidy, digits = 2)
#####

#####
# Count integer or numeric codes:
stat_sum2 <- epi_stats_summary(df_cont_chr,
                               codes = codes,
                               class_type = 'int_num',
                               action = 'codes_only'
                               )
stat_sum2
# Tidy and format them:
stat_sum_tidy <- epi_stats_tidy(sum_df = stat_sum2,
                                order_by = order_by,
                                perc_n = perc_n
                                )
stat_sum_tidy
epi_stats_format(stat_sum_tidy, digits = 0)
epi_stats_format(stat_sum_tidy, digits = 2, skip = c(2, 3))
#####

#####
# Get summary stats excluding contingency codes for character and factor columns:
stat_sum3 <- epi_stats_summary(df_cont_chr,
                               codes = codes,
                               class_type = 'chr_fct',
                               action = 'exclude'
                               )
stat_sum3
# Tidy and format:
stat_sum_tidy <- epi_stats_tidy(sum_df = stat_sum3,
                                order_by = order_by,
                                perc_n = perc_n
                                )
stat_sum_tidy
epi_stats_format(stat_sum_tidy, digits = 0)
epi_stats_format(stat_sum_tidy, digits = 1)
#####

#####
# Get summary stats for numeric/integer columns
# while excluding certain codes/values:
stat_sum4 <- epi_stats_summary(df = df_cont_chr,
                               codes = codes,
                               class_type = 'int_num',
                               action = 'exclude'
                               )
class(stat_sum4)
stat_sum4
# Numeric data summary doesn't need tidying but could be formatted:
epi_stats_format(stat_sum4, digits = 2)
#####

#####
# If there are no codes to return the result is an empty data.frame (tibble):
codes <- c('Per', 'X', '55')
stat_sum_zero <- epi_stats_summary(df_cont_chr,
                                   codes = codes,
                                   class_type = 'chr_fct',
                                   action = 'codes_only'
                                   )
stat_sum_zero
#####

AntonioJBT/episcout documentation built on Dec. 1, 2024, 4:07 a.m.