In Big-Life-Lab/bllflow: Big Life Lab Flow - a Workflow for Predictive Studies

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

bllflow builds from the tableone package to present the study cohort and description statistics. Also planned are tools to help create a study codebook.

Create "Table 1"

Create "Table 1- description of study data" for all variables in your database. This method is from the tableone package.

library(survival)
data(pbc)
library("tableone")
cat_vars <- c("status", "trt", "ascites", "hepato", "spiders", "edema", "stage")
create_table_one(data = pbc, factorVars = cat_vars)

Create Table 1 using the Model Specification Workbook

Create a Table 1 with only the variables in your model, using the Model Specification Workbook. First initialize the BLLFlow model.

library(bllflow)
variables_sheet <- read.csv(file.path(getwd(), '../inst/extdata/PBC-variables.csv'))
variables_details_sheet <- read.csv(file.path(getwd(), '../inst/extdata/PBC-variableDetails.csv'))
pbc_model <- BLLFlow(pbc, variables_sheet, variables_details_sheet)
create_table_one(pbc_model)

Create Table 1 with stratification

You can stratify Table 1 in two methods.

1) Stratify with columns. This is how to describe your data for manuscripts. This method is supported with the create_table_one library.

table_one <- create_table_one(data = pbc,strata = c("trt"), factorVars = cat_vars)

2) Stratify with rows. This method is helpful if there are many strata. We use this format for interactive visualizations or when we create figures. For example, see an 'algorithm viewer' that shows Table 1 stratified for 61 strata (bins).

table_one_long <- summary_data_long(table_one)

Add labels and metadata

For all tables, metadata such as labels are added from the Model Specification Workbook and/or DDI documents. Initialize the model with the DDI document.

ddi <- bllflow::ReadDDI(system.file("extdata", "", package = "bllflow"), "pbcDDI.xml")
pbc_model <- bllflow::update_msw(BLL_model = pbc_model, new_DDI = ddi) 
long_table_with_labels <- summary_data_long(table_one = table_one, bllflow_model = pbc_model, long_table = table_one_long)

Check for small cells

Our team works with personal health data in secure settings. For privacy, no summary tables can be exported from the data centre with small cells.

table_one <- create_table_one(data = pbc,strata = c("trt","stage"), factor_vars = cat_vars)
checked_table_one <- check_small_cells(table_one)

To obtain print of the actual small cells pass print as TRUE

checked_table_one <- check_small_cells(table_one, print = TRUE)

Find out which rows and variables contain the small cells.

checked_table_one$MetaData$small_cells

You can also check small cells inside your Summary Data in case you have multiple tables

checked_long_table <- check_small_cells(long_table_with_labels)

For a print of the found rows pass print as TRUE

checked_long_table <- check_small_cells(long_table_with_labels, print = TRUE)

Find out which rows and variables contain the small cells.

checked_long_table$MetaData$small_cells

Big-Life-Lab/bllflow documentation built on Feb. 1, 2023, 12:29 p.m.