sARDen

sARDen(SAS Analysis Results Display engine) is a SAS toolkit for producing CDISC ARS–aligned Analysis Results Data (ARD). Inspired by the R pharmaverse cards schema, it respectfully mirrors ARD conventions for cross-language workflows.
Macros cover summaries, tabulations, and hierarchical stacking, outputting long-format ARDs plus ARDM-like metadata via XATTR for traceability.

Test Data

We use ADSL and ADAE created as test data with the “sas_faker” package (https://github.com/Morioka-Yutaka/sas_faker),
but you can freely substitute typical ADSL/ADAE datasets instead, so feel free to adapt that part as you like.

%loadPackage(sas_faker)
%sas_faker(
n_groups=3, 
n_per_group=50,
output_lib=WORK,
seed =123456,
create_dm = N,
create_ae = N,
create_sv =  N,
create_vs = N,
create_adsl = Y,
create_adae = Y,
create_advs = N
);

[ADSL]

[ADAE]

`%sard_summary()` macro

Purpose:

Generate ARD-style summary statistics (long format) for continuous variables,
aligned to cards/ARS-like naming: group#, variable, context, stat_name, stat_label, stat, fmt_fun.
Can either compute statistics from raw data (data=) or post-process external stat data (statdata=).

Parameters:

  data=                 Source dataset for raw summaries (mutually exclusive with statdata=).  
  statdata=             Pre-computed statistics dataset (mutually exclusive with data=).  
  statdata_value=       Numeric value column in statdata to map into stat (default: COL1).  
  by=                   Grouping variables (space-separated). Must be non-empty for this version.  
  variable=             Analysis variables to summarize (space-separated). Required when data= is used.  
  statistic=            Statistics to keep (space-separated).  
                         Expected tokens (case-insensitive): N MEDIAN MEAN SD MIN MAX P25 P75 etc.  
  classdata=            Optional CLASSDATA= dataset for PROC SUMMARY to force level display.  
  context=              Context label stored in output (default: summary).  
  out=                  Output ARD dataset name.

Outputs:

  out= dataset with columns:  
    group1-groupN, group1_level-groupN_level, variable, context, stat_name, stat_label, stat, fmt_fun  
  Ordering variables_sort/statistic_sort used internally to preserve requested order.

Notes:

When data= is used, fmt_fun is auto-derived from decimal precision (capped at 3 dp).
When statdata= is used, fmt_fun is taken as-is; statdata must contain NAME in "var_stat" form.
stat_name "stddev" is normalized to "sd".

Example:

  %sard_summary(  
    data=ADSL,  
    by=TRT01P,  
    variable=AGE WEIGHTBL,  
    statistic=N MEDIAN MIN MAX MEAN SD,  
    out=sard_summary_mean  
  );

proc sort data=ADSL out=_ADSL;
   by TRT01P;
run;
proc surveymeans data=_ADSL geomean  gmstderr ;
   by TRT01P;
  var WEIGHTBL HEIGHTBL;
  ods output GeometricMeans=geomean_1;
run;
proc sort data=geomean_1;
  by  TRT01P VarName;
run;
proc transpose data=geomean_1 out=geomean_2(rename=(_NAME_=__NAME_)) prefix=value;
  by TRT01P VarName;
run;
data geomean_3;
set geomean_2;
_NAME_=catx("_",varname,__NAME_);
fmt_fun ="2";
run;
%sard_summary(
  statdata = geomean_3,
  statdata_value = value1,
  by =TRT01P,
  variable = WEIGHTBL HEIGHTBL,
  statistic = GeoMean GMStdErr,
  out = sard_summary_geomean
)

`%sard_tabulate()` macro

Purpose:

Generate ARD-style tabulations (long format) for categorical variables,
producing n, BigN (denominator), and p (proportion), compatible with cards tabulate ARD.

Parameters:

  data=                   Source dataset.  
  variable=               Categorical variables to tabulate (space-separated). Required.  
  by=                     Optional grouping variables (space-separated).  
  statistic=              Stats to output (space-separated). Default: n P BigN.  
                           n    = count within variable level  
                           BigN = denominator for the by-group  
                           p    = n/BigN  
  denominator_dataset=    Optional dataset to define denominators explicitly.  
                           If blank, denominators are derived from data=.  
  classdata=              Optional CLASSDATA= for PROC SUMMARY to force level display.  
  context=                Context label stored in output (default: tabulate).  
  out=                    Output ARD dataset name.

Outputs:

out= dataset with columns:
group1-groupN, group1_level-groupN_level (if by provided),
variable, variable_level, context, stat_name, stat_label, stat, fmt_fun
fmt_fun:
n/BigN -> "0", p -> "xx.x" (intended display template).

Notes:

If by= is empty, denominators are scalar and applied to all levels.
variable_level stored as formatted value (vvalue).
Assumes variable/bystats exist and are non-missing for numerator records.

Example:

  %sard_tabulate(  
    data=ADSL,  
    by=TRT01P,  
    variable=SEX,  
    statistic=n P BigN,  
    out=sard_tabulate  
  );

`%sard_stack_hierarchical()` macro

Purpose:

Generate hierarchical ARD for rate-style stacking (subject-level unique counting).
Designed for event hierarchies like AE SOC/PT, producing overall + per-level summaries by group.

Parameters:

  data=                   Source event dataset (e.g., ADAE).  
  variable=               Hierarchical variables (space-separated), ordered top->bottom (e.g., AEBODSYS AEDECOD).  
  variable_hieral_code=   Optional ordering code variables aligned with variable= list (space-separated).  
  by=                     Grouping variables (space-separated), e.g., TRTA.  
  id=                     Subject identifier for rate counting (unique per id within each hierarchy level). Required.  
  statistic=              Stats to output (space-separated). Default: n P BigN.  
  denominator_dataset=    Dataset to define denominators (usually ADSL). Required for p/BigN usage.  
  classdata=              Optional CLASSDATA= for PROC SUMMARY.  
  over_variables=         Y/N.  
                           Y -> include overall ("Any event") row via dummy variable.  
                           N -> no overall row.  
  out=                    Output ARD dataset name.

Outputs:

out= dataset with columns:
group#, group#_level, variable, variable_level, context="hierarchical",
stat_name, stat_label, stat, fmt_fun, plus optional variable_hieral_code columns.
Includes overall row where variable="hierarchical_overall", variable_level="Y" if over_variables=Y.

Notes:

Subject-level de-duplication is enforced per (id, by, variable level).
variable_hieral_code assumes same length as variable= and key uniqueness in data=.
by= must be non-empty for this version.

Example:

  %sard_stack_hierarchical(  
    data=ADAE,  
    variable=AEBODSYS AEDECOD,  
    variable_hieral_code=AEBDSYCD F_AEPTCD,  
    by=TRTA,  
    id=USUBJID,  
    denominator_dataset=ADSL(rename=(TRT01A=TRTA)),  
    out=sard_stack_hierarchical  
  );

`%sard_stack_hierarchical_count()` macro

Purpose:

Generate hierarchical ARD for count-style stacking (record/event counting).
Similar to sard_stack_hierarchical, but does not de-duplicate by subject id.

Parameters:

  data=                   Source event dataset (e.g., ADAE).  
  variable=               Hierarchical variables (space-separated), ordered top->bottom.  
  variable_hieral_code=   Optional ordering code variables aligned with variable= list.  
  by=                     Grouping variables (space-separated).  
  statistic=              Stats to output (space-separated). Default: n.  
  denominator_dataset=    Dataset to define denominators (optional unless p/BigN requested).  
  classdata=              Optional CLASSDATA= for PROC SUMMARY.  
  over_variables=         Y/N to include overall dummy row.  
  out=                    Output ARD dataset name.

Outputs:

out= dataset with columns:
group#, group#_level, variable, variable_level, context="hierarchical",
stat_name, stat_label, stat, fmt_fun, plus optional variable_hieral_code columns.

Notes:

Counts reflect number of records/events, not number of subjects.
Denominator/pct layers are computed even if statistic excludes them (harmless overhead).
variable_hieral_code assumes same length as variable=.

Example:

  %sard_stack_hierarchical_count(  
    data=ADAE,  
    variable=AEBODSYS AEDECOD,  
    variable_hieral_code=AEBDSYCD F_AEPTCD,  
    by=TRTA,  
    statistic=n,  
    denominator_dataset=ADSL(rename=(TRT01A=TRTA)),  
    out=sard_stack_hierarchical_count  
  );

`%bind_ard()` macro

Purpose:

Bind multiple ARD datasets (long format) into a single ARD,
optionally removing duplicates and normalizing column order to ARD-friendly sequence.

parameters:

  indata=          Space-separated list of ARD datasets to stack.  
  outdata=         Output dataset name.  
  distinct=        Y/N.  
                   Y -> remove duplicate rows across inputs (excluding dsno/record_no).  
                   N -> keep all rows.  
  drop_sortno=     Y/N.  
                   Y -> drop internal dsno and record_no from output.  
                   N -> keep them.

Outputs:

outdata= dataset containing stacked ARD rows.
Column order is set to: group# / group#_level / variable / variable_level / context /
stat_name / stat_label / stat / fmt_fun (any other columns follow).

Notes:

dsno increments when CUROBS resets between input datasets.
distinct=Y assumes compatible column structure across ARDs to avoid unintended row drops.
Current version uses "informat &vsort;" as a placeholder; column order relies on DICTIONARY sequence.

Example:

  %bind_ard(  
    indata=Sard_summary_mean Sard_tabulate Sard_stack_hierarchical_count,  
    outdata=ard_tab_14_2_1,  
    distinct=Y,  
    drop_sortno=Y  
  );

`%xrdm_set()` macro

Purpose:

Attach ARDM/ARM-TS-like metadata to an ARD dataset using SAS extended attributes (XATTR).
Supports partial updates; only non-empty parameters are written.

Parameters:

  lib=              Library where ARD dataset resides (default: WORK).  
  ard=              ARD dataset name to modify. Required.  
  result_id=        Result identifier (e.g., R001).  
  analysis_id=      Analysis identifier (e.g., A001).  
  method_id=        Method identifier (e.g., M001).  
  result_context=   Free-text description of result context.  
  table_id=         Table/TFL identifier (e.g., T14.2.1).  
  display_label=    Free-text display label for output.  
  print=            Y/N.  
                    Y -> print ExtendedAttributesDS section via PROC CONTENTS after update.  
                    N -> no print.

Outputs:

Updates dataset-level extended attributes on &lib..&ard.
Optionally prints attribute listing to log/output.

Dependencies / Side Effects:

Uses PROC DATASETS MODIFY with XATTR SET DS.
Does not create intermediate datasets.

Notes:

Free-text fields should be passed with %nrbquote(...) if they contain special characters.
Existing keys are overwritten when re-specified.

Example:

  %xrdm_set(  
    lib=WORK,  
    ard=ard_tab_14_2_1,  
    result_id=R001,  
    analysis_id=A001,  
    method_id=M001,  
    result_context=%nrbquote(Drug vs Placebo on CHG),  
    table_id=%nrbquote(T14.2.1),  
    display_label=%nrbquote(Mean Difference (Drug–Placebo))  
  );

version history

0.1.0(01December2025): Initial version

What is SAS Packages?

The package is built on top of SAS Packages Framework(SPF) developed by Bartosz Jablonski.

For more information about the framework, see SAS Packages Framework.

You can also find more SAS Packages (SASPacs) in the SAS Packages Archive(SASPAC).

How to use SAS Packages? (quick start)

1. Set-up SAS Packages Framework

First, create a directory for your packages and assign a packages fileref to it.

filename packages "\path\to\your\packages";

Secondly, enable the SAS Packages Framework. (If you don't have SAS Packages Framework installed, follow the instruction in SPF documentation to install SAS Packages Framework.)

%include packages(SPFinit.sas)

2. Install SAS package

Install SAS package you want to use with the SPF's %installPackage() macro.

For packages located in SAS Packages Archive(SASPAC) run:
```
%installPackage(packageName)
```

For packages located in PharmaForest run:

%installPackage(packageName, mirror=PharmaForest)

For packages located at some network location run:
```
%installPackage(packageName, sourcePath=https://some/internet/location/for/packages)
```
(e.g. %installPackage(ABC, sourcePath=https://github.com/SomeRepo/ABC/raw/main/))

3. Load SAS package

Load SAS package you want to use with the SPF's %loadPackage() macro.

%loadPackage(packageName)

Name		Name	Last commit message	Last commit date
Latest commit History 23 Commits
hist		hist
sARDen/06_macro		sARDen/06_macro
LICENSE		LICENSE
README.md		README.md
description.sas		description.sas
license.sas		license.sas
sARDen.png		sARDen.png
sARDen_small.png		sARDen_small.png
sarden.md		sarden.md
sarden.zip		sarden.zip

Folders and files

Latest commit

History

Repository files navigation

sARDen

Test Data

%sard_summary() macro

Purpose:

Parameters:

Outputs:

Notes:

Example:

%sard_tabulate() macro

Purpose:

Parameters:

Outputs:

Notes:

Example:

%sard_stack_hierarchical() macro

Purpose:

Parameters:

Outputs:

Notes:

Example:

%sard_stack_hierarchical_count() macro

Purpose:

Parameters:

Outputs:

Notes:

Example:

%bind_ard() macro

Purpose:

parameters:

Outputs:

Notes:

Example:

%xrdm_set() macro

Purpose:

Parameters:

Outputs:

Dependencies / Side Effects:

Notes:

Example:

version history

What is SAS Packages?

How to use SAS Packages? (quick start)

1. Set-up SAS Packages Framework

2. Install SAS package

3. Load SAS package

Enjoy!

About

Resources

License

Licenses found

Uh oh!

Stars

Watchers

Forks

Releases

Packages 0

Uh oh!

Contributors

Uh oh!

Languages

`%sard_summary()` macro

`%sard_tabulate()` macro

`%sard_stack_hierarchical()` macro

`%sard_stack_hierarchical_count()` macro

`%bind_ard()` macro

`%xrdm_set()` macro

Packages