1.4. Migrating from SAS to R: A Skill Conversion Guide

1.4.16. The apply Family Functions in R vs SAS

1. Introduction

The apply family of functions in R (apply, lapply, sapply, vapply, mapply, tapply, etc.) provides a powerful and concise way to perform operations on data structures without explicit loops. These functions are essential for efficient data manipulation, especially in clinical trial programming, and offer a flexible alternative to SAS's array processing and summary procedures.

2. SAS vs R: Program Comparison

2A. Table: Common Use Cases for apply Family Functions

3. apply(): Row/Column Operations on Matrices and Data Frames

Description:
apply() is used to apply a function to the rows or columns of a matrix or numeric data frame. It is especially useful for summarizing repeated measures (e.g., labs, vitals) for each subject.

Possible Uses:

• Calculate row or column means, sums, min, max, standard deviation, etc.
• Summarize repeated measures for each subject (e.g., multiple lab results, vital signs).
• Apply custom functions to each row or column.

Input Table (Lab Results - LB):

• The second argument in apply function MARGIN, determines whether the function should be applied across rows or columns:
• MARGIN = 1 → apply the function across rows (i.e., row-wise)
• MARGIN = 2 → apply the function across columns (i.e., column-wise)

R Row Example:

# Dummy LB data
lb <- data.frame(
  USUBJID = c("01-001", "01-002", "01-003"),
  LBTESTCD = c("HGB", "HGB", "HGB"),
  LBORRES1 = c(13.2, 12.8, 14.0),
  LBORRES2 = c(13.5, 13.0, 13.8),
  LBORRES3 = c(13.1, 12.9, 14.2)
)
lb$row_mean <- apply(lb[, c("LBORRES1", "LBORRES2", "LBORRES3")], 1, mean)
lb$row_sd <- apply(lb[, c("LBORRES1", "LBORRES2", "LBORRES3")], 1, sd)
lb$row_min <- apply(lb[, c("LBORRES1", "LBORRES2", "LBORRES3")], 1, min)
lb$row_max <- apply(lb[, c("LBORRES1", "LBORRES2", "LBORRES3")], 1, max)

Output Table:

R Column Example: We’ll now apply a function column-wise to the three lab result columns.

apply(lb[, c("LBORRES1", "LBORRES2", "LBORRES3")], 2, mean)

Explanation

• lb[, c("LBORRES1", "LBORRES2", "LBORRES3")]: selects the numeric lab result columns.
• 2: tells apply() to operate column-wise.
• mean: calculates the mean of each column.

Output

LBORRES1 LBORRES2 LBORRES3 
   13.33     13.43     13.40 

This tells us:

• The average of LBORRES1 across all subjects is 13.33
• The average of LBORRES2 is 13.43
• The average of LBORRES3 is 13.40

4. lapply(): Element-wise Operations on Lists or Data Frame Columns

Description:
lapply() applies a function to each element of a list or each column of a data frame, always returning a list. It is commonly used for cleaning or transforming all columns of a domain.

Possible Uses:

• Apply a function to each column of a data frame (e.g., uppercase, trim whitespace).
• Apply a function to each element of a list (e.g., summary, length, custom transformation).
• Convert all factors to characters or vice versa.

Input Table (Demographics - DM):

R Example:

# Dummy DM data
dm <- data.frame(
  USUBJID = c("01-001", "01-002", "01-003"),
  SEX = c("F", "M", "F"),
  RACE = c("Asian", "White", "Black"),
  stringsAsFactors = FALSE
)
char_cols <- sapply(dm, is.character)
dm[char_cols] <- lapply(dm[char_cols], toupper)

Output Table:

5. sapply(): Simplified Element-wise Operations

Description:
sapply() is similar to lapply() but tries to simplify the result to a vector or matrix. It is useful for getting summary statistics or properties for each element/column.

Possible Uses:

• Get the length, class, or summary of each column.
• Apply a function to each element and return a vector.
• Quick checks or summaries for reporting.

Input Table (Adverse Events - AE):

R Example:

# Dummy AE data
ae <- data.frame(AEDECOD = c("HEADACHE", "NAUSEA", "DIZZINESS"))
sapply(ae$AEDECOD, nchar)

Output:

6. tapply(): Grouped Summaries

Description:
tapply() applies a function over subsets of a vector, defined by a factor (grouping variable). It is ideal for grouped summaries.

Possible Uses:

• Calculate mean, sum, min, max by group (e.g., mean AGE by SEX, mean lab by visit).
• Summarize adverse events by severity or seriousness.
• Count occurrences by group.

Input Table:

R Example:

dm <- data.frame(
  USUBJID = c("01-001", "01-002", "01-003"),
  SEX = c("F", "M", "F"),
  AGE = c(34, 40, 29)
)
tapply(dm$AGE, dm$SEX, mean)

Output:

Example: Group by Multiple Columns with tapply()

To convert the tapply() output (which is typically a matrix or array) into a data frame, you can use the as.data.frame() function combined with reshape techniques. Here's how you can do it using the SDTM-style VS domain example: Example: tapply() → Data Frame

# Sample SDTM-style data
vs <- data.frame(
  USUBJID = c("SUBJ001", "SUBJ001", "SUBJ002", "SUBJ002", "SUBJ003", "SUBJ003"),
  VSTESTCD = c("SYSBP", "DIABP", "SYSBP", "DIABP", "SYSBP", "DIABP"),
  VSSTRESN = c(120, 80, 130, 85, 125, 82),
  VISIT = c("SCREENING", "SCREENING", "SCREENING", "SCREENING", "SCREENING", "SCREENING")
)

# Group by USUBJID and VSTESTCD, compute mean
result <- tapply(vs$VSSTRESN, list(vs$USUBJID, vs$VSTESTCD), mean)

# Convert to data frame
df_result <- as.data.frame(as.table(result))
colnames(df_result) <- c("USUBJID", "VSTESTCD", "MEAN_VSSTRESN")

df_result

Output:

Explanation

• tapply(...): Computes the grouped mean.
• as.table(...): Converts the array to a table object.
• as.data.frame(...): Converts the table to a tidy data frame.
• colnames(...): Renames columns for clarity.

This is a clean and efficient way to get grouped summary statistics in a data frame format using only base R.

7. mapply(): Parallel Operations Across Multiple Vectors

Description:
mapply() applies a function in parallel over multiple arguments (vectors/lists). It is useful for row-wise operations combining multiple columns.

Possible Uses:

• Concatenate or combine values from multiple columns for reporting.
• Create custom flags or labels using multiple variables.
• Apply a function to corresponding elements of several vectors.

Input Table:

R Example:

ae <- data.frame(
  AEDECOD = c("HEADACHE", "NAUSEA", "DIZZINESS"),
  AESEV = c("MILD", "MODERATE", "SEVERE")
)
mapply(function(a, b) paste(a, b, sep = " - "), ae$AEDECOD, ae$AESEV)

Output:

8. replicate(): Simulations and Repeated Calculations

Description:
replicate() repeats an expression multiple times, useful for simulations or generating mock data.

Possible Uses:

• Simulate random subject ages or lab values.
• Bootstrap resampling.
• Generate mock data for testing.

R Example:

set.seed(123)
replicate(5, sample(20:70, 1))

Output:
A vector of 5 random ages, e.g., [38 70 57 21 67]

9. calling Custom Function

We'll use a simulated VS (Vital Signs) domain and apply a custom function using the apply family.

Step 1: Simulate an SDTM-like VS Dataset

vs <- data.frame(
  USUBJID = rep(c("SUBJ001", "SUBJ002", "SUBJ003"), each = 4),
  VSTESTCD = rep(c("SYSBP", "DIABP"), times = 6),
  VISIT = rep(c("SCREENING", "WEEK1"), times = 6),
  VSSTRESN = c(120, 80, 130, 85, 125, 82, 118, 78, 135, 88, 140, 90)
)

Step 2: Define a Custom Function

cv <- function(x) {
  if (mean(x) == 0) return(NA)
  return(sd(x) / mean(x))
}

Step 3: Apply the Custom Function Using tapply()

tapply(vs$VSSTRESN, list(vs$VSTESTCD, vs$VISIT), cv)

Output:

         SCREENING     WEEK1
DIABP   0.03608439 0.06681531
SYSBP   0.04166667 0.07856742

Step 4: Convert to Data Frame (Optional)

cv_result <- as.data.frame(as.table(
  tapply(vs$VSSTRESN, list(vs$VSTESTCD, vs$VISIT), cv)
))
colnames(cv_result) <- c("VSTESTCD", "VISIT", "CV")

**Resource download links**

1.4.16.-The-apply-Family-Functions-in-R-vs-SAS.zip