Comprehensive Comparison: Missing Values Support

SAS vs. Stata vs. SPSS - Detailed Analysis

SAS 9.4 M9 / Viya 2025.07

Stata 19 (April 2025)

SPSS 31 (June 2025)

Executive Summary

This document provides a detailed comparison of missing values support across three major statistical software packages. The analysis identifies identical features, key differences, and unique capabilities of each platform.

Key Finding: SAS and Stata share a categorical/alphabetic approach with 28 and 27 missing value types respectively, while SPSS uses a fundamentally different value-based definition approach.

Quick Reference Comparison

Feature	SAS	Stata	SPSS
Latest Version	9.4 M9 / Viya 2025.07	19 (April 2025)	31 (June 2025)
Numeric Missing Types	28 total	27 total	System + User-defined
Extended/Special Missing	Yes (27: ._ + .A-.Z)	Yes (26: .a-.z)	No (different approach)
Character/String Missing Types	1 (blank only)	1 (blank only)	Up to 3 discrete values
Range-Based Missing	No	No	Yes (numeric only)
Can Label System Missing	Yes	No	N/A
Case Sensitivity (coding)	No	No	N/A

1. Numeric Variables - Missing Values

1.1 Total Number of Missing Value Types

Software	Total Types	Breakdown	Notes
SAS	28	1 standard (.) + 1 special (._) + 26 letters (.A-.Z)	Most options
Stata	27	1 system (.) + 26 letters (.a-.z)	Very similar to SAS
SPSS	Variable	1 system + user-defined (3 discrete OR 1 range + 1 discrete)	Fundamentally different approach

IDENTICAL:

All three support the concept of numeric missing values
All automatically exclude missing values from statistical calculations

DIFFERENT:

SAS and Stata use categorical/alphabetic system; SPSS uses value-based definition
SAS has 28 types; Stata has 27 types; SPSS has flexible user-defined approach
SAS includes ._ (underscore) as lowest missing; Stata and SPSS do not

1.2 Extended/Special Missing Values

Software	Support	Syntax	Display	Purpose
SAS	✓ Yes	`._, .A-.Z` (or `.a-.z`)	Upper case	Categorical reasons for missingness
Stata	✓ Yes	`.a-.z` (or `.A-.Z`)	Lower case	Categorical reasons for missingness
SPSS	✗ No	N/A	N/A	Uses value-based missing instead

IDENTICAL (SAS & Stata):

Both support alphabetic extended/special missing values
Both allow 26 letter-based missing values
Both are case-insensitive in code

DIFFERENT:

SAS has additional ._ (underscore) as 28th missing value
SAS displays uppercase; Stata displays lowercase
SPSS has no extended missing concept - uses completely different approach

1.3 Range-Based Missing Values

Software	Support	Example Syntax	Use Case
SAS	✗ No	N/A	Cannot define range as missing
Stata	✗ No	N/A	Cannot define range as missing
SPSS	✓ Yes	`MISSING VALUES income (-99 THRU -1).`	Define continuous range as missing

IDENTICAL (SAS & Stata): Neither supports range-based missing values

DIFFERENT:

SPSS uniquely supports defining ranges (e.g., -99 through -1) as missing
SPSS can combine: 1 range + 1 additional discrete value
SAS and Stata require explicit coding for each missing type

1.4 Missing Value Sort Order

Software	Sort Order	Comparison Behavior
SAS	`._ < . < .A < .B < ... < .Z`	All missing > any non-missing number
Stata	`. < .a < .b < ... < .z`	All missing > any non-missing number
SPSS	System missing has special handling	System missing ≠ specific value

2. Character/String Variables - Missing Values

2.1 Number of Missing Value Types

Software	Missing Types	Representation	Specification
SAS	1	Blank: `' '` or `" "`	Single space between quotes
Stata	1	Blank: `""`	Empty string
SPSS	Up to 3	Discrete values	User specifies up to 3 distinct strings

IDENTICAL (SAS & Stata):

Both support only 1 character/string missing value (blank)
All three: Blank/empty string can represent missing

DIFFERENT:

SPSS allows up to 3 discrete string values as missing (most flexible)
SAS and Stata limited to blank only
SPSS example: MISSING VALUES gender ('X', 'NK', '').

3. Value Labels / Formats

3.1 Labeling Mechanism

Software	Mechanism	Syntax Pattern	Terminology
SAS	PROC FORMAT	`proc format; value name val='label'; run;`	Formats
Stata	label define	`label define name val "label"`	Value labels
SPSS	VALUE LABELS	`VALUE LABELS var val 'label'.`	Value labels

3.2 Labeling Missing Values

Software	Can Label Missing?	Which Missing?	Limitations
SAS	✓ Yes	All types (., ._, .A-.Z)	None - complete flexibility
Stata	~ Partial	Extended only (.a-.z)	Cannot label system missing (.)
SPSS	✓ Yes	User-defined missing values	Only user-defined, not system missing

KEY DIFFERENCE:

SAS: Can label ALL missing types including standard (.)
Stata: Cannot label system missing (.), only extended (.a-.z)
SPSS: Can label user-defined missing values
Winner: SAS most flexible for missing value labeling

3.3 String/Character Variable Labels

Software	Direct Labeling	Approach
SAS	✓ Yes	`value $formatname` ($ prefix for character)
Stata	✗ No	Must encode string to numeric first
SPSS	✓ Yes	Direct VALUE LABELS application to strings

4. Code Examples: Defining Missing Values with Labels

4.1 Complete Code Examples - All Possible Missing Values

The following table shows complete, executable code examples for defining all possible missing values with labels in each software package. Each example demonstrates the maximum capabilities of the system.

Missing Value Type	SAS Code Example	Stata Code Example	SPSS Code Example
NUMERIC VARIABLES - DEFINING MISSING VALUES
Standard/System Missing	`/* Automatic - always exists */ age = .;`	`// Automatic - always exists age = .`	`* Automatic - always exists COMPUTE age = $SYSMIS.`
Special Missing: Refused	`age = .A; /* or .a */`	`replace age = .a`	`* Use user-defined value COMPUTE age = -9. MISSING VALUES age (-9).`
Special Missing: Not Applicable	`age = .B; /* or .b */`	`replace age = .b`	`* Use user-defined value COMPUTE age = -8. MISSING VALUES age (-9, -8).`
Special Missing: Don't Know	`age = .C; /* or .c */`	`replace age = .c`	`* Use user-defined value COMPUTE age = -7. MISSING VALUES age (-9, -8, -7).`
All Extended/Special Missing (.A-.Z / .a-.z)	`/* SAS supports 28 total: . (standard) ._ (underscore - lowest) .A, .B, .C, ... .Z (26 letters) / age = ._; / Lowest missing / age = .A; / to / age = .Z; / 26 letter options */`	`// Stata supports 27 total: // . (system) // .a, .b, .c, ... .z (26 letters) replace age = . // System replace age = .a // to replace age = .z // 26 letter options`	`* SPSS: Up to 3 discrete OR * 1 range + 1 discrete * Option 1: Three discrete values MISSING VALUES age (-9, -8, -7). * Option 2: One range + one value MISSING VALUES age (-99 THRU -1, 999).`
NUMERIC VARIABLES - LABELING MISSING VALUES
Complete Format/Label Definition	`proc format; value agefmt 18-25 = 'Young Adult' 26-40 = 'Adult' 41-65 = 'Middle Age' 66-high = 'Senior' . = 'Unknown' ._ = 'Lowest Missing' .A = 'Refused' .B = 'Not Applicable' .C = 'Don''t Know' .D = 'Not Reported' .E = 'Other Missing' .F-.Z = 'Other Special'; run; /* Apply format */ format age agefmt.;`	`label define age_lbl /// 18 "18 years" /// 25 "25 years" /// 35 "35 years" /// .a "Refused" /// .b "Not Applicable" /// .c "Don't Know" /// .d "Not Reported" /// .e "Other Missing" /// .f "Invalid" /// .z "End Missing", modify // NOTE: System missing (.) // cannot be labeled in Stata label values age age_lbl`	`* Define user-defined missing MISSING VALUES age (-9, -8, -7). * Label all values including missing VALUE LABELS age 18 '18 years' 25 '25 years' 35 '35 years' -9 'Refused' -8 'Not Applicable' -7 'Don''t Know'. * Note: System missing cannot * be labeled in SPSS`
CHARACTER/STRING VARIABLES - DEFINING MISSING VALUES
String Missing: Blank	`gender = ' '; /* or " " / / Blank is the only character missing */`	`replace gender = "" // Blank is the only // string missing`	`COMPUTE gender = ''. MISSING VALUES gender ('').`
String Missing: Multiple Values	`/* NOT SUPPORTED Only blank ' ' is missing for character variables */`	`// NOT SUPPORTED // Only blank "" is missing // for string variables`	`* SPSS supports up to 3 discrete MISSING VALUES gender ('X', 'NK', '').`
CHARACTER/STRING VARIABLES - LABELING VALUES
Complete String Format/Label	`proc format; value $genderfmt 'M' = 'Male' 'F' = 'Female' 'X' = 'Other' ' ' = 'Not Reported' other = 'Unknown'; run; format gender $genderfmt.;`	`// Stata requires encoding // strings to use labels encode gender, /// gen(gender_num) /// label(gender_lbl) // Then modify the label label define gender_lbl /// 1 "Male" /// 2 "Female" /// 3 "Other", modify`	`* Direct labeling supported VALUE LABELS gender 'M' 'Male' 'F' 'Female' 'X' 'Other' '' 'Not Reported'.`
COMPLETE WORKING EXAMPLE - ALL FEATURES
Full Working Example	`/* Create dataset / data demo; age = 25; output; age = .A; output; / Refused / age = .B; output; / N/A / age = .; output; / Unknown / run; / Define formats / proc format; value agefmt low-17 = 'Under 18' 18-65 = 'Adult' 66-high = 'Senior' . = 'Unknown' .A = 'Refused' .B = 'Not Applicable' .C-.Z = 'Other Missing'; run; / Apply and display */ data demo; set demo; format age agefmt.; run; proc freq data=demo; tables age / missing; run;`	`// Create dataset clear set obs 4 gen age = . replace age = 25 in 1 replace age = .a in 2 // Refused replace age = .b in 3 // N/A replace age = . in 4 // Unknown // Define labels label define age_lbl /// 18 "18 years" /// 25 "25 years" /// 65 "65 years" /// .a "Refused" /// .b "Not Applicable" /// .c "Don't Know" /// .d-.z "Other Missing", modify // Apply labels label values age age_lbl // Display with missing tabulate age, missing`	`* Create dataset DATA LIST FREE / age. BEGIN DATA 25 -9 -8 . END DATA. * Define missing values MISSING VALUES age (-9, -8). * Label all values VALUE LABELS age 18 '18 years' 25 '25 years' 65 '65 years' -9 'Refused' -8 'Not Applicable'. * Display frequencies FREQUENCIES VARIABLES=age /ORDER=ANALYSIS.`

Key Observations from Code Examples:

SAS: Most comprehensive - 28 missing types, can label ALL including standard (.)
Stata: 27 missing types, cannot label system missing (.), requires encoding for string labels
SPSS: Different paradigm - user defines which values are missing (flexible but requires documentation)
String Variables: Only SPSS supports multiple discrete string missing values (up to 3)
Ranges: Only SPSS supports range-based missing (e.g., -99 THRU -1)

5. Detection and Testing

5.1 Functions for Missing Detection

Software	Function/Test	Detects	Notes
SAS	`missing(var)`	All 28 types	Universal check
	`var = .`	Standard (.) only	Specific check
	`var <= .Z`	All missing	Range check
Stata	`missing(var)`	All 27 types	Universal check
	`var == .`	System (.) only	Specific check
	`var < .`	Exclude all missing	Comparison check
SPSS	`MISSING(var)`	System + user-defined	Universal check
	`SYSMIS(var)`	System missing only	Specific check
	`VALUE(var)`	User-defined missing	Specific check

5.2 Code Examples for Detection

Detection Task	SAS Code	Stata Code	SPSS Code
Check if any missing	`if missing(age) then flag = 1;`	`gen flag = missing(age)`	`IF MISSING(age) flag=1.`
Check for system/standard missing only	`if age = . then flag_std = 1;`	`gen flag_std = (age == .)`	`IF SYSMIS(age) flag_std=1.`
Check for specific special missing	`if age = .A then flag_refused = 1;`	`gen flag_refused = (age == .a)`	`IF (age = -9) flag_refused=1.`
Exclude all missing in condition	`if age <= .Z then delete; /* Or */ if age < . then process;`	`drop if missing(age) // Or keep non-missing keep if age < .`	`SELECT IF NOT MISSING(age).`
Count missing by type	`data counts; set mydata; if age = . then cnt_std + 1; if age = .A then cnt_ref + 1; if age = .B then cnt_na + 1; run;`	`count if age == . count if age == .a count if age == .b // Or use tabulate tab age, missing`	`FREQUENCIES age /MISSING.`

IDENTICAL:

All three: Provide universal function to detect any missing type
All three: missing() or MISSING() function available
All three: Can test for specific missing types

5. Handling in Statistical Procedures

5.1 Automatic Exclusion

Software	Automatic Exclusion?	User Control
SAS	✓ Yes	Procedures exclude all missing by default
Stata	✓ Yes	Commands exclude all missing by default
SPSS	✓ Yes	Procedures exclude all missing by default

IDENTICAL:

All three: Automatically exclude missing values from statistical calculations
All three: Provide options to include missing in tabulations
All three: Consistent behavior across procedures

Summary: Identical vs. Different Features

Features IDENTICAL Across All Three

Feature	SAS	Stata	SPSS
Concept of missing values for numeric	✓	✓	✓
Concept of missing values for character/string	✓	✓	✓
Automatic exclusion from statistics	✓	✓	✓
Can label/format regular values	✓	✓	✓
Functions to detect missing	✓	✓	✓
Options to include missing in tables	✓	✓	✓
Stable across recent versions	✓	✓	✓

Unique Features by Software

SAS Unique Features

28 missing values - Most missing value types (includes ._)
Can label standard missing - Only one that can label . (dot)
Character formats ($) - Direct character variable formats
Complete flexibility - Label all missing types without exception

Stata Unique Features

27 missing values - Extended missing (.a-.z)
misstable command - Dedicated missing pattern analysis suite
Clean notation - No ._ to remember, simpler than SAS
Lowercase display - Consistent lowercase output

SPSS Unique Features

Range-based missing - Only one supporting ranges (e.g., -99 THRU -1)
3 string missing values - Most flexible for character variables
Direct string labels - Can apply VALUE LABELS directly to strings
User-friendly approach - Value-based definition more intuitive

Key Differences Summary

Aspect	SAS	Stata	SPSS
Missing Value Approach	Categorical (28 types)	Categorical (27 types)	Value-based (user-defined)
String Missing Flexibility	Low (1 type)	Low (1 type)	High (3 types)
Range Support	No	No	Yes
Labeling Completeness	Complete (all types)	Partial (not .)	User-defined only
Ease of Use (beginners)	Moderate	Moderate	High
Power (advanced users)	High	High	Moderate

Recommendations by Use Case

Choose SAS if:

Need maximum missing value types (28)
Want to label all missing types including standard
Complex surveys with many non-response categories
Need most comprehensive missing value system

Choose Stata if:

Need extended missing values (27 is usually sufficient)
Want dedicated missing pattern analysis (misstable)
Prefer simpler notation (no ._ to remember)
Similar power to SAS with slightly cleaner syntax

Choose SPSS if:

Working with pre-coded missing values (999, -9, etc.)
Need range-based missing (all negative = missing)
Need multiple discrete string missing values
Want most intuitive interface for beginners
Prefer value-based rather than categorical approach

Conclusion

Overall Assessment

IDENTICAL CORE CONCEPTS: All three software packages understand and handle missing values, automatically exclude them from analysis, provide detection functions, and maintain stable implementations across versions.

KEY PHILOSOPHICAL DIVIDE:

SAS and Stata: Share nearly identical categorical/alphabetic approach (28 vs 27 types) - asking "WHY is data missing?"
SPSS: Uses fundamentally different value-based definition approach - asking "WHICH values are missing?"

FLEXIBILITY RANKING:

SAS: Most missing types (28), can label all types, complete flexibility
Stata: Nearly as powerful (27 types), excellent tools (misstable)
SPSS: Different strengths (ranges, multiple string missing), more user-friendly

BEST PRACTICE: The choice depends on data source (pre-coded missing vs. new collection), complexity needs (simple vs. many missing types), user expertise (beginners vs. advanced), and organizational standards and existing workflows.

All three are fully capable statistical packages with robust missing value support - the differences are in philosophy and implementation details rather than fundamental capability.