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

1.4.11. IF/ELSE statements in SAS vs R equivalent

1. Basic Conditional Logic: Simple Value Transformation

SAS Example

data ae_severity;
  set ae;
  length severity_cat $20 urgency_level 8;
  if aesev = 'MILD' then severity_cat = 'Non-significant';
  else if aesev = 'MODERATE' then severity_cat = 'Notable';
  else if aesev = 'SEVERE' then severity_cat = 'Significant';
  else if aesev = 'LIFE THREATENING' then severity_cat = 'Critical';

  if aesev = 'MILD' then urgency_level = 1;
  else if aesev = 'MODERATE' then urgency_level = 2;
  else if aesev = 'SEVERE' then urgency_level = 3;
  else if aesev = 'LIFE THREATENING' then urgency_level = 4;
run;

Explanation (SAS):

• Uses if/else if statements to categorize adverse events based on severity (AESEV)
• Creates both character (severity_cat) and numeric (urgency_level) variables
• Each severity level gets a corresponding category and numeric urgency level
• Sequential logic ensures each adverse event falls into only one category

R Example

library(dplyr)

# Dummy data for AE example
ae <- data.frame(
  USUBJID = c("001-001", "001-002", "001-003", "001-004"),
  AEDECOD = c("HEADACHE", "NAUSEA", "DYSPNEA", "ANAPHYLAXIS"),
  AESEV = c("MILD", "MODERATE", "SEVERE", "LIFE THREATENING")
)

ae_severity <- ae %>%
  mutate(
    SEVERITY_CAT = case_when(
      AESEV == "MILD" ~ "Non-significant",
      AESEV == "MODERATE" ~ "Notable",
      AESEV == "SEVERE" ~ "Significant",
      AESEV == "LIFE THREATENING" ~ "Critical"
    ),
    URGENCY_LEVEL = case_when(
      AESEV == "MILD" ~ 1,
      AESEV == "MODERATE" ~ 2,
      AESEV == "SEVERE" ~ 3,
      AESEV == "LIFE THREATENING" ~ 4
    )
  )

Explanation (R):

• Uses case_when() function from dplyr package
• Conditions appear on the left side of ~, return values on the right
• Multiple conditions evaluated in order until one is TRUE
• Multiple variables can be created in a single mutate() call
• case_when() handles each column separately (no direct equivalent to DO blocks)

Input Table

Expected Output

2. Compound Conditions: Multiple Criteria

SAS Example

data vs_category;
  set vs;
  if vstestcd = 'SYSBP' & vsstresn > 180 then clinical_flag = 'Critical Hypertension';
  else if vstestcd = 'SYSBP' & vsstresn > 140 then clinical_flag = 'Hypertension';
  else if vstestcd = 'DIABP' & vsstresn > 120 then clinical_flag = 'Critical Diastolic';
  else if vstestcd = 'TEMP' & vsstresn > 38.5 then clinical_flag = 'Fever';
  else clinical_flag = 'Normal Range';
run;

Explanation (SAS):

• Combines multiple conditions (vital sign test code and result value)
• Uses logical AND (&) to ensure both conditions must be met
• Classifies vital signs measurements based on clinical thresholds
• Creates a flagging variable for abnormal values requiring attention

R Example

library(dplyr)

# Dummy data for VS example
vs <- data.frame(
  USUBJID = c("001-001", "001-002", "001-003", "001-004", "001-005"),
  VSTESTCD = c("SYSBP", "SYSBP", "DIABP", "TEMP", "PULSE"),
  VSSTRESN = c(185, 145, 125, 39.2, 88)
)

vs_category <- vs %>%
  mutate(
    CLINICAL_FLAG = case_when(
      VSTESTCD == "SYSBP" & VSSTRESN > 180 ~ "Critical Hypertension",
      VSTESTCD == "SYSBP" & VSSTRESN > 140 ~ "Hypertension",
      VSTESTCD == "DIABP" & VSSTRESN > 120 ~ "Critical Diastolic",
      VSTESTCD == "TEMP" & VSSTRESN > 38.5 ~ "Fever",
      TRUE ~ "Normal Range"
    )
  )

Explanation (R):

• Combines conditions with & (AND) operator, just like in SAS
• Values that don't match any condition become NA by default
• The case_when() function processes conditions in order
• Each observation gets assigned the first matching condition's value

Input Table

Expected Output

3. Default Values: Handling Unmatched Conditions

SAS Example

data cm_importance;
  set cm;
  if cmclas = 'ACE INHIBITORS' & cmstrf = 'Y' then priority = 'Critical';
  else if cmclas = 'ACE INHIBITORS' then priority = 'High';
  else if cmstrf = 'Y' then priority = 'Medium';
  else if cmstdy < 0 then priority = 'Baseline';
  else priority = 'Standard';
run;

Explanation (SAS):

• Categorizes concomitant medications based on drug class and study relevance flag
• Adds a final else statement as a catch-all for medications not meeting special criteria
• All medications are guaranteed to have a priority assigned
• The catch-all is only executed if all previous conditions are FALSE

R Example

library(dplyr)

# Dummy data for CM example
cm <- data.frame(
  USUBJID = c("001-001", "001-002", "001-003", "001-004", "001-005"),
  CMCLAS = c("ACE INHIBITORS", "ACE INHIBITORS", "ANALGESICS", "ANALGESICS", "ANTIBIOTICS"),
  CMSTRF = c("Y", "N", "Y", "N", "N"),
  CMSTDY = c(10, 5, -3, -5, 15)
)

cm_importance <- cm %>%
  mutate(
    PRIORITY = case_when(
      CMCLAS == "ACE INHIBITORS" & CMSTRF == "Y" ~ "Critical",
      CMCLAS == "ACE INHIBITORS" ~ "High",
      CMSTRF == "Y" ~ "Medium",
      CMSTDY < 0 ~ "Baseline",
      TRUE ~ "Standard"
    )
  )

Explanation (R):

• Uses TRUE ~ as the final condition to catch all remaining cases
• This works because TRUE is always TRUE for any observation
• Equivalent to the else statement in SAS
• Ensures all observations get a value (no NAs for unmatched conditions)

Input Table

Expected Output

4. Multiple Actions: DO Blocks and Equivalent R Approaches

SAS Example

data ex_derived;
  set ex;
  if exdose > 100 & exdosu = 'mg' then do;
    dose_category = 'High';
    monitoring_freq = 'Daily';
    max_duration = 14;
  end;
  else if exdose > 50 & exdosu = 'mg' then do;
    dose_category = 'Medium';
    monitoring_freq = 'Weekly';
    max_duration = 28;
  end;
  else if exdose > 0 then do;
    dose_category = 'Low';
    monitoring_freq = 'Monthly';
    max_duration = 90;
  end;
  else do;
    dose_category = 'Unknown';
    monitoring_freq = 'Unknown';
    max_duration = 0;
  end;
  
  /* Calculate additional derived values */
  total_dose = exdose * exdur;
  avg_daily = total_dose / exdur;
run;

Explanation (SAS):

• Uses DO blocks to perform multiple related actions for each condition
• For each order, assigns appropriate discount, shipping cost, and delivery time
• Final calculations use the values assigned in the conditional blocks
• Clear visual grouping of related operations per condition

R Example

library(dplyr)

# Dummy data for EX example
ex <- data.frame(
  USUBJID = c("001-001", "001-002", "001-003", "001-004"),
  EXTRT = c("DRUG-A", "DRUG-A", "DRUG-B", "PLACEBO"),
  EXDOSE = c(120, 75, 25, 0),
  EXDOSU = c("mg", "mg", "mg", "mg"),
  EXDUR = c(7, 14, 30, 30)
)

ex_derived <- ex %>%
  mutate(
    DOSE_CATEGORY = case_when(
      EXDOSE > 100 & EXDOSU == "mg" ~ "High",
      EXDOSE > 50 & EXDOSU == "mg" ~ "Medium",
      EXDOSE > 0 ~ "Low",
      TRUE ~ "Unknown"
    ),
    MONITORING_FREQ = case_when(
      EXDOSE > 100 & EXDOSU == "mg" ~ "Daily",
      EXDOSE > 50 & EXDOSU == "mg" ~ "Weekly",
      EXDOSE > 0 ~ "Monthly",
      TRUE ~ "Unknown"
    ),
    MAX_DURATION = case_when(
      EXDOSE > 100 & EXDOSU == "mg" ~ 14,
      EXDOSE > 50 & EXDOSU == "mg" ~ 28,
      EXDOSE > 0 ~ 90,
      TRUE ~ 0
    )
  ) %>%
  mutate(
    TOTAL_DOSE = EXDOSE * EXDUR,
    AVG_DAILY = TOTAL_DOSE / EXDUR
  )

Explanation (R):

• No direct equivalent to DO blocks in dplyr's approach
• Instead, repeat the conditions for each variable being modified
• Each column transformation uses its own case_when() function
• R's approach is more column-oriented, while SAS is row-oriented

Input Table

Expected Output

5. Handling Missing Values: NA Conditions

SAS Example

data qs_complete;
  set qs;
  if qsstresn = 5 then response_level = "Complete Response";
  else if qsstresn = 4 then response_level = "Strong Response";
  else if qsstresn = 3 then response_level = "Moderate Response";
  else if qsstresn = 2 then response_level = "Mild Response";
  else if qsstresn = 1 then response_level = "No Response";
  else if qsstresn = . then response_level = "Missing Response";
run;

Explanation (SAS):

• Handles survey responses with satisfaction scores from 1-5
• Special condition for missing scores (satisfaction_score = .)
• Categorizes responses into descriptive result categories
• Explicitly handles missing values with a specific condition

R Example - Approach 1: Using case_when() with is.na()

library(dplyr)

# Dummy data for QS example
qs <- data.frame(
  USUBJID = c("001-001", "001-002", "001-003", "001-004", "001-005"),
  QSTEST = rep("PAIN ASSESSMENT", 5),
  QSSTRESN = c(5, 3, 1, NA, 4)
)

qs_complete1 <- qs %>%
  mutate(
    RESPONSE_LEVEL = case_when(
      QSSTRESN == 5 ~ "Complete Response",
      QSSTRESN == 4 ~ "Strong Response",
      QSSTRESN == 3 ~ "Moderate Response",
      QSSTRESN == 2 ~ "Mild Response",
      QSSTRESN == 1 ~ "No Response",
      is.na(QSSTRESN) ~ "Missing Response"
    )
  )

Explanation (R - Approach 1):

• Uses is.na() function to explicitly test for NA values
• Special handling required as case_when() silently skips NA inputs
• Order matters: test for exact match, non-missing non-match, then missing
• !is.na() ensures we don't incorrectly evaluate NA values

R Example - Approach 2: Using if_else() chain

library(dplyr)

# Dummy data for QS example (already defined above)
qs_complete2 <- qs %>%
  mutate(
    RESPONSE_LEVEL = if_else(QSSTRESN == 5, "Complete Response",
                    if_else(QSSTRESN == 4, "Strong Response",
                    if_else(QSSTRESN == 3, "Moderate Response",
                    if_else(QSSTRESN == 2, "Mild Response",
                    if_else(QSSTRESN == 1, "No Response",
                          "Missing Response")))))
  )

Explanation (R - Approach 2):

• if_else() provides a simpler syntax for simple conditions
• Takes condition, true value, false value, and missing value
• More concise than case_when() for simple binary conditions
• The missing parameter specifies what to return for NA inputs

Input Table

Expected Output

6. Beyond Basics: Complex Conditional Logic

SAS Example

data lb_flagging;
  set lb;
  
  /* Direct lab abnormality flagging based on reference ranges */
  if lbstresn > (lbstnrhi * 3) & lbstnrhi > 0 then do;
    abnormality = "Critically High";
    flag = "H++";
    action = "Repeat Test Immediately";
    score = 3;
  end;
  else if lbstresn > lbstnrhi & lbstnrhi > 0 then do;
    abnormality = "High";
    flag = "H";
    action = "Monitor";
    score = 1;
  end;
  else if lbstresn < (lbstnrlo * 0.5) & lbstnrlo > 0 then do;
    abnormality = "Critically Low";
    flag = "L++";
    action = "Repeat Test Immediately";
    score = 3;
  end;
  else if lbstresn < lbstnrlo & lbstnrlo > 0 then do;
    abnormality = "Low";
    flag = "L";
    action = "Monitor";
    score = 1;
  end;
  else do;
    abnormality = "Normal";
    flag = "N";
    action = "No Action";
    score = 0;
  end;
  
  /* Additional processing for specific tests */
  if lbtestcd = "ALT" then do;
    if lbstresn > (lbstnrhi * 3) then priority = "Hepatic Alert";
    else if lbstresn > lbstnrhi then priority = "Hepatic Monitor";
    else priority = "Routine";
  end;
  else priority = "Standard";
run;

Explanation (SAS):

• Directly compares lab results against reference ranges
• Uses simple multipliers for critical thresholds (3x upper limit, 0.5x lower limit)
• Assigns a simplified score based on severity level
• Includes special processing for liver function tests (ALT)
• Avoids unnecessary intermediate calculations

R Example

library(dplyr)

# Dummy data for LB example
lb <- data.frame(
  USUBJID = c("001-001", "001-002", "001-003", "001-004", "001-005"),
  LBTESTCD = c("ALT", "AST", "HGB", "PLT", "ALT"),
  LBSTRESN = c(150, 55, 9, 140, 25),
  LBSTNRLO = c(10, 10, 12, 150, 10),
  LBSTNRHI = c(40, 40, 16, 450, 40)
)

lb_flagging <- lb %>%
  mutate(
    # Direct lab abnormality flagging based on reference ranges
    ABNORMALITY = case_when(
      LBSTRESN > (LBSTNRHI * 3) & LBSTNRHI > 0 ~ "Critically High",
      LBSTRESN > LBSTNRHI & LBSTNRHI > 0 ~ "High",
      LBSTRESN < (LBSTNRLO * 0.5) & LBSTNRLO > 0 ~ "Critically Low",
      LBSTRESN < LBSTNRLO & LBSTNRLO > 0 ~ "Low",
      TRUE ~ "Normal"
    ),
    FLAG = case_when(
      LBSTRESN > (LBSTNRHI * 3) & LBSTNRHI > 0 ~ "H++",
      LBSTRESN > LBSTNRHI & LBSTNRHI > 0 ~ "H",
      LBSTRESN < (LBSTNRLO * 0.5) & LBSTNRLO > 0 ~ "L++",
      LBSTRESN < LBSTNRLO & LBSTNRLO > 0 ~ "L",
      TRUE ~ "N"
    ),
    ACTION = case_when(
      LBSTRESN > (LBSTNRHI * 3) & LBSTNRHI > 0 ~ "Repeat Test Immediately",
      LBSTRESN < (LBSTNRLO * 0.5) & LBSTNRLO > 0 ~ "Repeat Test Immediately",
      LBSTRESN > LBSTNRHI & LBSTNRHI > 0 ~ "Monitor",
      LBSTRESN < LBSTNRLO & LBSTNRLO > 0 ~ "Monitor",
      TRUE ~ "No Action"
    ),
    SCORE = case_when(
      LBSTRESN > (LBSTNRHI * 3) & LBSTNRHI > 0 ~ 3,
      LBSTRESN < (LBSTNRLO * 0.5) & LBSTNRLO > 0 ~ 3,
      LBSTRESN > LBSTNRHI & LBSTNRHI > 0 ~ 1,
      LBSTRESN < LBSTNRLO & LBSTNRLO > 0 ~ 1,
      TRUE ~ 0
    )
  ) %>%
  # Additional processing for specific tests
  mutate(
    PRIORITY = case_when(
      LBTESTCD == "ALT" & LBSTRESN > (LBSTNRHI * 3) ~ "Hepatic Alert",
      LBTESTCD == "ALT" & LBSTRESN > LBSTNRHI ~ "Hepatic Monitor",
      LBTESTCD == "ALT" ~ "Routine",
      TRUE ~ "Standard"
    )
  )

Explanation (R):

• Uses direct comparisons with reference ranges rather than calculating intermediate ratios
• Simplified scoring with straightforward severity levels (3 = critical, 1 = abnormal, 0 = normal)
• Maintains the same logical structure and outcomes as the original example
• Makes the code more readable by removing unnecessary complexity
• Still allows for special handling of specific test types (ALT)

Input Table

Expected Output

7. Beyond Basics: Vectorized Alternatives in R

SAS Example

/* Standard IF/ELSE approach */
data ds_status1;
  set ds;
  if dsdecod = 'COMPLETED' then status = 'Completed';
  else if dsdecod = 'ADVERSE EVENT' then status = 'Discontinued Due to AE';
  else if dsdecod = 'LACK OF EFFICACY' then status = 'Discontinued Due to LOE';
  else if dsdecod = 'WITHDRAWAL BY SUBJECT' then status = 'Withdrawn';
  else if dsdecod = 'LOST TO FOLLOW-UP' then status = 'LTFU';
  else status = 'Other Discontinuation';
run;

/* FORMAT approach as alternative */
proc format;
  value $dsstatfmt 'COMPLETED' = 'Completed'
                   'ADVERSE EVENT' = 'Discontinued Due to AE'
                   'LACK OF EFFICACY' = 'Discontinued Due to LOE'
                   'WITHDRAWAL BY SUBJECT' = 'Withdrawn'
                   'LOST TO FOLLOW-UP' = 'LTFU'
                   other = 'Other Discontinuation';
run;

data ds_status2;
  set ds;
  status = put(dsdecod, $dsstatfmt.);
run;

Explanation (SAS):

• IF/ELSE is the standard approach for conditional logic
• FORMAT approach can be an alternative for simple value mapping
• Less flexible but potentially more concise for straight value mapping
• Both approaches produce the same result

R Example - Multiple Approaches

library(dplyr)

# Dummy data for DS example
ds <- data.frame(
  USUBJID = c("001-001", "001-002", "001-003", "001-004", "001-005"),
  DSTERM = c("COMPLETED STUDY", "DISCONTINUED DUE TO AE", "LACK OF EFFICACY", "PATIENT WITHDREW", "RELOCATION"),
  DSDECOD = c("COMPLETED", "ADVERSE EVENT", "LACK OF EFFICACY", "WITHDRAWAL BY SUBJECT", "OTHER")
)

# Standard case_when approach
ds_status1 <- ds %>%
  mutate(
    STATUS = case_when(
      DSDECOD == "COMPLETED" ~ "Completed",
      DSDECOD == "ADVERSE EVENT" ~ "Discontinued Due to AE",
      DSDECOD == "LACK OF EFFICACY" ~ "Discontinued Due to LOE",
      DSDECOD == "WITHDRAWAL BY SUBJECT" ~ "Withdrawn",
      DSDECOD == "LOST TO FOLLOW-UP" ~ "LTFU",
      TRUE ~ "Other Discontinuation"
    )
  )

# Alternative with recode
ds_status2 <- ds %>%
  mutate(
    STATUS = recode(DSDECOD,
      "COMPLETED" = "Completed",
      "ADVERSE EVENT" = "Discontinued Due to AE",
      "LACK OF EFFICACY" = "Discontinued Due to LOE",
      "WITHDRAWAL BY SUBJECT" = "Withdrawn",
      "LOST TO FOLLOW-UP" = "LTFU",
      .default = "Other Discontinuation"
    )
  )

# Alternative using named vector lookup
status_values <- c(
  "COMPLETED" = "Completed",
  "ADVERSE EVENT" = "Discontinued Due to AE",
  "LACK OF EFFICACY" = "Discontinued Due to LOE",
  "WITHDRAWAL BY SUBJECT" = "Withdrawn",
  "LOST TO FOLLOW-UP" = "LTFU"
)

ds_status3 <- ds %>%
  mutate(
    STATUS = ifelse(DSDECOD %in% names(status_values),
                   status_values[DSDECOD],
                   "Other Discontinuation")
  )

Explanation (R):

• R offers multiple approaches for conditional transformations
• case_when(): Most similar to IF/ELSE, best for complex conditions
• recode(): Concise for simple value mapping (like SAS FORMATs)
• ifelse(): Can be nested for multiple conditions
• Each approach has its own use cases and advantages

Input Table

Expected Output (all approaches)

8. Summary: SAS vs R Conditional Logic Capabilities

Key Points:

• SAS IF/ELSE provides a procedural, row-by-row approach that's intuitive for complex logic
• R's case_when() offers a vectorized approach that's more column-oriented
• SAS DO blocks have no direct equivalent in R's tidyverse approach
• R offers more alternatives for simple value mapping (recode, etc.)
• Both systems require special handling for missing values
• SAS is often more readable for complex nested logic
• R's pipe operator (%>%) helps chain operations for complex transformations
• For simple value mapping, R's recode() is similar to SAS formats

When to use each approach:

• SAS IF/ELSE: When you need complex row-based operations with multiple actions per condition
• R case_when(): For most conditional operations, especially with multiple outcomes
• R if_else(): For simple binary conditions, especially with missing value handling
• R recode(): For simple value mapping/recoding with no complex conditions

**Resource download links**

1.4.11.-IF-ELSE-statements-in-SAS-vs-R-equivalent.zip