← SASnRMerging datasets - cartesian product

Introduction

A Cartesian product (cross join) pairs every row of one dataset with every row of another, producing every possible combination.
In clinical programming this is handy for building summary shells: row categories crossed with treatment arms, for example.
SAS does this with a DATA step using two SET statements without a BY, or via PROC SQL.
Tidyverse provides cross_join(). Base R uses merge() with by = NULL.
This lesson crosses three summary row templates with four treatment arms to produce 12 combinations.
After running, the output dataset has rows * columns = 3 * 4 = 12 rows.

Merging datasets - cartesian product

SAS (Base SAS)+

R (tidyverse)+

library(tidyverse)

dummy01<-tribble(
~grouplabel,~statistic,~group,~intord,
"Sex, n(%)","Male",2,1,
"Sex, n(%)","Female",2,2,
"Sex, n(%)","Missing",2,99,
)

dummy_treatments<-tribble(
~treatment,
1,
2,
3,
4,
)

dummy02<-cross_join(dummy01,dummy_treatments)

In tidyverse, cross_join() from dplyr builds the full cross-product in a single call.
cross_join(dummy01, dummy_treatments) pairs every row of the first with every row of the second — n1 * n2 rows in the result.
No by = argument is needed — cross_join deliberately ignores keys.
For cross-joining named vectors rather than tibbles, the same result comes from tidyr::crossing().
The expected row count is always n1 * n2, which we check with nrow() before moving on.

Happy learning!

Search lessons

Introduction

Merging datasets - cartesian product