In jjwillard/wfbmcphsr: WFBMC Public Health Sciences

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>"
)

In health science research, producing tables of baseline characteristics (commonly called Table 1) of patient cohorts is a routine procedure. Much of the analysis-specific data cleaning work is required in the steps leading up to table formation, but without a streamlined process for manipulating and organizing values of resulting calculations, considerable time can be spent pasting results together. The make_table_one function attempts to make this calculation, manipulation, and organization of summary values as streamlined as possible.

Overview

Pre-function data cleaning:

• Import your data
• Clean appropriately. Note that for binary variables, you should recode your levels to 0-1 values as results will show the summary values of level 1 only. (Follow standard naming convention, where 'YES' = 1 and 'NO' = 0)
• Create vectors of names for numeric variables, binary variables, multiple level categorical variables, and vectors corresponding to subgroups of interest for which you would like to calculate the mean or counts of another variable (see below)
• Make all categorical variables factors (this includes binary, multiple level categorical and subgroups)
• Finally, create a vector containing the order of variables as you wish them to be displayed in your final table

Using make_table_one:

• Substitute in named vectors in appropriate spots in the make_table_one function
• Results are returned invisibly so you should save them as an object to view them
• There is also an option to export the sorted data table as an RTF for further editing/formatting

The function make_table_one

Usage

make_table_one(df, grouping_var, num_vars = NULL, num_display = "PM",
  binary_cat_vars = NULL, multiple_cat_vars = NULL, cat_display = "CP", 
  subgroups_m = NULL, mean_vars_for_subgroups = NULL, 
  subgroups_c = NULL, count_vars_for_subgroups = NULL, 
  order_of_vars = NULL, 
  export_rtf = FALSE, rtf_filename = NULL,
  show_pval = TRUE, digits = 2)

Parameters

• df: Dataset containing covariates
• grouping_var: Variable to group by (will be columns of table). Must be included (ie. one column summary tables currently not allowed).
• num_vars: Vector of numeric variables
• num_display: How should numeric variable results be displayed? ('PM' for mean +- sd, 'PRS' for mean (sd), defaults to 'PM')
• binary_cat_vars: Vector of binary variables (in 0-1 format)
• multiple_cat_vars: Vector of multiple level categorical variables
• cat_display: How should categorical results be displayed? ('CP' = counts and proportions, 'C' = counts, 'P' = proportions, defaults to 'CP')
• subgroups_m: Vector of subgroup variables (must be factors) of interest for calculating means of a numeric variable. NOTE: If you plan to analyze a categorical variable by itself and then also use it for subgroup analysis, it is advised to copy and rename the columns of the categorical variables you wish to use as subgroups (ie. rename variable to something like 'variable_sub_m") so Table One can be properly sorted to reflect your specification in 'order_of_vars'. If the subgroup does not have a unique name, it will be sorted directly below the categorical variable.
• mean_vars_for_subgroups: Vector of numeric variables from which to calculate mean and sd (must match vector position of corresponding subgroups_m variable)
• subgroups_c: Vector of subgroup variables (must be factors) of interest for calculating counts of a BINARY variable. NOTE: Follow same naming convention of subgroups as that listed in subgroups_m.
• count_vars_for_subgroups: Vector of BINARY variables from which to calculate mean and sd (must match vector position of corresponding subgroups_c variable.) Results are displayed for the case where the binary count variable is equal to 1.
• order_of_vars: A vector of all variables listed in the order of the rows you desire for your Table One.
• export_rtf: Logical. Do you want to export an RTF to location specified in 'rtf_filename'? (Defaults to FALSE)
• rtf_filename: Must be quoted and include '.rtf' extension
• show_pval: Logical. Should the p-value results be displayed? Defaults to TRUE.
• digits: Number of digits to round decimals. Defaults to 2.

Pre-function data cleaning

Once you have imported your data, clean it appropriately:

baseline <- read_sas("my_sas_file.sas7bdat")

my_cleaned_baseline <- baseline %>%
  mutate(var1 = if_else(ex_var1, 0, 1),
     var2 = if_else(ex_var1, 0, 1),
     var3 = if_else(ex_var3 == 1, 1, 0)) %>%
  select(grouping_var, some_vector_of_other_variables)

Now make vectors of the names of your different types of variables. NOTE: If you plan to analyze a categorical variable by itself and then also use it for subgroup analysis, it is advised to copy and rename the columns of the categorical variables you wish to use as subgroups (ie. rename variable to something like 'variable_sub") so make_table_one can be properly sorted to reflect your specification in 'order_of_vars'. If the subgroup does not have a unique name, it will be sorted directly below the categorical variable.

num_vars <- c('num_var1', 'numvar2', ...)
binary_cat_vars <- c('bin_var1', 'bin_var2', ...)
multiple_cat_vars <- c('mc_var1', 'mc_var2', ...)

subgroups_m <- c('subgroup_m_1', subgroup_m_2', ...)
mean_vars <- c('calulate_mean_of_this_for_subgroup_m_1', 'calulate_mean_of_this_for_subgroup_m_2', ...)

subgroups_c <- c('subgroup_c_1', subgroup_c_2', ...)
count_vars <- c('calulate_count_of_this_for_subgroup_c_1', 'calulate_count_of_this_for_subgroup_c_2', ...))

Convert your categorical variables into factors (VERY IMPORTANT):

my_cleaned_baseline[binary_cat_vars] <- lapply(my_cleaned_baseline[binary_cat_vars], factor)
my_cleaned_baseline[multiple_cat_vars] <- lapply(my_cleaned_baseline[multiple_cat_vars], factor)
my_cleaned_baseline[subgroups_m] <- lapply(my_cleaned_baseline[subgroups_m], factor)
my_cleaned_baseline[subgroups_c] <- lapply(my_cleaned_baseline[subgroups_c], factor)

Create a vector containing the order of variables as you wish them to be displayed in your final table:

var_order <- c('var_for_first_row', 'var_for_second_row', ...)

Calling make_table_one

Once you have your data cleaned and vectors of variable names as above, substitute into make_table_one and save as an object for viewing within R.

table_1 <- make_table_one(df = my_cleaned_baseline, grouping_var = random_assignment_group, 
  num_vars = num_vars, num_display = 'PM', 
  binary_cat_vars = binary_cat_vars, multiple_cat_vars = multiple_cat_vars, cat_display = 'CP',
  subgroups_m = subgroups_m, mean_vars_for_subgroups = mean_vars, 
  subgroups_c = subgroups_c, count_vars_for_subgroups = count_vars,
  order_of_vars = var_order,
  export_rtf = TRUE, rtf_filename = 'my_location/my_filename.rtf', 
  show_pval = TRUE, digits = 2)

Notice that export_rtf = TRUE so an RTF will be generated at the location specified in rtf_filename = 'my_location/my_filename.rtf'.

Future Work

In the future, I hope to add the following:

• More error handling to deal with common type and value issues
• A simulated baseline dataset to play with in the package
• A full tutorial with the simulated dataset walking through each step
• Any other functionality you think might be convenient to have (email me!)

Thanks for using make_table_one!

jjwillard/wfbmcphsr documentation built on May 14, 2019, 5:01 a.m.