Abstract:
|
Tabular, descriptive summaries of data are a core component of reporting in many contexts, both regulatory and non-regulatory. Often, a single table can represent tabulations (or more complex analyses) on multiple, nested levels of subsetting on a single dataset. In this case, rows may not be independent, but rather represent different levels of detail - e.g., summaries of age: across all patients, within different ethnicities, and for each reported gender within each of those ethnicities. Furthermore, columns often represent an orthogonal axis of subsetting on the data - e.g., by trial arm or even baseline vs follow-up nested within trial arm. With rtables, we present a pipe-able grammar for declaring descriptive summary and analysis tables based on multi-stage, nested splitting of raw input data. By declaring a tabulation before splitting the data, we are able to both automate the data subsetting process and produce table layouts that can be applied to many similarly-structured datasets. This approach is particularly powerful in regulatory or similar settings where data can be expected to follow a formal specification.
|