← SASnRFind Maximum Value Across Multiple Variables in Each Row

Introduction

Clinical data often spreads repeated measures across columns — hgb_v1, hgb_v2, hgb_v3 — and a derivation needs the per-row maximum across those columns.
The answer is a new column whose value on each row is the largest of the listed columns, ignoring missings.
SAS handles this in a DATA step with the function max(of var1--varN), where the of keyword signals a variable list.
Tidyverse uses rowwise() + mutate(max = max(...)) to evaluate the max one row at a time.
Base R uses pmax() (parallel max across vectors) or apply(..., 1, max) on selected columns.
This lesson picks the maximum hemoglobin across three visits, so each subject ends up with a single maximum value.

Find Maximum Value Across Multiple Variables in Each Row

SAS (Base SAS)+

data lab;
input usubjid hgb_v1 hgb_v2 hgb_v3;
datalines;
1001 12.5 13.1 12.8
1002 13.0 14.0 13.5
1003 11.2 11.8 12.0
;
run;

data lab02;
   set lab;
   maxval = max(of hgb_v1--hgb_v3);
run;

In SAS, row-wise aggregation across columns uses functions with the of keyword followed by a variable list.
maxval = max(of hgb_v1--hgb_v3); computes the maximum of the three columns on each row.
The double-dash (--) expands to every variable between hgb_v1 and hgb_v3 in column order — a concise list when the columns sit next to each other.
If any column is missing, SAS silently skips it in the max() call; a row with all three missing returns missing.
We can swap max for min, mean, sum, or n to compute any of the common row-wise summaries the same way.
We run and check the maxval column matches the largest of the three input columns on each row.

R (tidyverse)+

lab <- tribble(
  ~USUBJID, ~HGB_V1, ~HGB_V2, ~HGB_V3,
  1001,     12.5,     13.1,    12.8,
  1002,     13.0,     14,      13.5,
  1003,     11.2,     11.8,    12.0
)

lab01 <- lab %>% 
  mutate(
    maxval=max(HGB_V1,HGB_V2,HGB_V3)
  )

lab02 <- lab %>% 
  rowwise() %>% 
  mutate(
    maxval=max(HGB_V1,HGB_V2,HGB_V3)
  )

In tidyverse, mutate() works column-wise by default, so a naive max(HGB_V1, HGB_V2, HGB_V3) returns one value across the whole tibble, not one value per row.
The row-wise form uses rowwise() before mutate() — dplyr then evaluates the call one row at a time.
lab %>% rowwise() %>% mutate(maxval = max(HGB_V1, HGB_V2, HGB_V3)) gives the correct per-row maximum.
We remember to call ungroup() afterwards so the tibble goes back to ordinary (column-wise) behaviour for downstream steps.
An alternative that avoids rowwise(): mutate(maxval = pmax(HGB_V1, HGB_V2, HGB_V3)) — pmax() is vectorised across the inputs.
For many columns, use mutate(maxval = pmax(!!!across(starts_with("HGB_")))) or the modern across() with .fns = list(max = max).
We run and compare lab01$maxval (wrong, single value) against lab02$maxval (correct per-row) to see the difference.

Happy learning!

Search lessons

Introduction

Find Maximum Value Across Multiple Variables in Each Row