Day 3 of 30·R Basics & SAS Equivalents

Numeric Functions: ROUND, INT, SUM and the Missing-Value Trap

Numeric functions translate cleanly — but R handles missing values differently, and that difference matters for every derivation.

Learning objectives

1Map common SAS numeric functions (ROUND, INT, CEIL, MEAN, SUM) to R.
2Understand how NA propagates and why na.rm = TRUE is essential.
3Contrast SAS's MEAN-of-nonmissing default with R's strict NA behavior.
4Perform safe rounding and integer arithmetic for clinical derivations.

Key SAS^® → R comparison

Task	SAS^®	R
Round	ROUND(x, 0.1)	round(x, 1)
Truncate to integer	INT(x)	trunc(x)
Ceiling / floor	CEIL(x) / FLOOR(x)	ceiling(x) / floor(x)
Absolute value	ABS(x)	abs(x)
Square root	SQRT(x)	sqrt(x)
Sum across values	SUM(a, b, c) (skips missing)	sum(a, b, c, na.rm = TRUE)
Mean	MEAN(of x1-x5) (skips missing)	mean(x, na.rm = TRUE)
Max / Min	MAX(a,b) / MIN(a,b)	max(a,b, na.rm=TRUE) / min(...)
Is missing?	if x = . then ...	if (is.na(x)) ...

Code examples

Rounding a derived BMI

SAS®

data vs;
  set vs;
  bmi = round(weight / (height/100)**2, 0.1);
run;

vs <- vs |>
  mutate(bmi = round(weight / (height/100)^2, 1))

Note: round(x, 1) rounds to one decimal place. Note R uses ^ for exponentiation where SAS uses **.

The missing-value trap

SAS®

data lb;
  set lb;
  /* SAS SUM/MEAN automatically skip missing */
  total = sum(of result1-result3);
  avg   = mean(of result1-result3);
run;

lb <- lb |>
  mutate(
    total = rowSums(across(result1:result3), na.rm = TRUE),
    avg   = rowMeans(across(result1:result3), na.rm = TRUE)
  )

# Plain sum() WITHOUT na.rm returns NA if any value is missing:
sum(c(1, 2, NA))            # NA
sum(c(1, 2, NA), na.rm = TRUE)  # 3

Note: This is the #1 gotcha for SAS programmers. SAS's SUM()/MEAN() silently ignore missing values; R's sum()/mean() return NA unless you add na.rm = TRUE.

Conditional logic on missing values

SAS®

data lb;
  set lb;
  if aval = . then flag = 'Missing';
  else if aval > uln then flag = 'High';
  else flag = 'Normal';
run;

lb <- lb |>
  mutate(flag = case_when(
    is.na(aval)   ~ "Missing",
    aval > uln    ~ "High",
    TRUE          ~ "Normal"
  ))

Note: Always test is.na() FIRST. In R, NA > uln evaluates to NA (not FALSE), so an untested missing value can slip through your conditions.

Explanation

Numeric functions are largely a vocabulary swap — round, abs, sqrt, ceiling, floor all behave as you'd expect. The one conceptual difference that will bite every new R programmer is missing-value handling. SAS treats the missing value as a special low value and its aggregate functions (SUM, MEAN, MIN, MAX) quietly drop missings. R treats NA as 'unknown', and the philosophy is that any computation involving an unknown is itself unknown — so mean(c(1,2,NA)) is NA, not 1.5.

The fix is the na.rm = TRUE argument, available on virtually every aggregation function. Get in the habit of adding it deliberately, and document when you do, because in clinical derivations the decision to include or exclude missing values is a scientific one, not a coding afterthought.

Be equally careful in conditional logic. Because comparisons against NA return NA rather than FALSE, an if/case_when that doesn't test is.na() first can misclassify records. The idiom is always to handle the missing case explicitly and first.

Practice exercise

Create a lab dataset with some missing AVAL values. Derive a change-from-baseline column, a per-subject mean using na.rm = TRUE, and a flag column using case_when() that returns 'Missing', 'Low', 'Normal', or 'High'. Then deliberately omit na.rm and is.na() to observe how NA propagates.

Learning objectives

Key SAS® → R comparison

Code examples

Explanation

Practice exercise

Key SAS^® → R comparison