PROC COMPARE is often seen as the gold standard for validating datasets within clinical trials. However the output it produces can often be extensive and manually interpreting this can be timely and error prone. This paper discusses an automated approach for examining the results of PROC COMPARE, by interpreting the return code that the procedure creates. This approach allows the user to quickly create single page summaries displaying the comparison results of multiple (i.e. more than 30) comparisons in an easy to interpret one observation per comparison format.

We have previously seen several papers that have explained how to utilise PROC COMPARE return codes, however what differentiates this paper is its clever use of the SAS bAND function to interpret the return codes.

The paper was presented at the SAS  Global Forum, 2012. Download the paper here.

The post SAS paper review 6-Deciphering PROC COMPARE return codes/ appeared first on Statskom.

The paper examines problems which can result from processing and storing large SDTM  datasets and how to resolve them. Although this paper was written in 2010, we feel that the issues and techniques discussed in this paper are particularly important in the light of the guidance provided by CDISC  (SDTM Implementation Guide 3.2 – section 4.1.2.9) that:

Very large transport files have become an issue for FDA to process. One of the main contributors to the large file sizes has been sponsors using the maximum length of 200 for character variables. To help rectify this situation:
• The maximum SAS Version 5 character variable length of 200 characters should not be used unless necessary.

The paper presents several methods for handling large datasets more efficiently and suggests ways in which the size of datasets can be reduced. We particularly liked the tables which show real time, CPU time and memory costs associated with different approaches, which provide a tangible illustration of the benefits of efficiency savings.

The paper was presented at the SESUG conference, Savannah, GA, 2010. Download the paper here.

The post SAS-paper-review-5/ appeared first on Statskom.

The paper looks at the problems which can be caused when non-printable and special characters such as line feeds and carriage returns are present in SAS datasets, how to detect them and how to replace the.

Non printable and special characters can be a source of confusion when they are encountered in SAS, the programmer often only detects them by chance when they notice that the alignment in an output report seems to be disrupted for no apparent reason. A systematic approach to detecting and handling special characters is clearly required in situations when special characters can be reasonably expected and this is just what is set out in this paper.

The paper describes a clear approach for detecting special characters and explains three keyapproaches for dealing with them:

1. Report them to data management and get clean data in the next transfer.
2. Replace NPSC with other characters.
3. Delete NPSC.

Particularly useful is the recommendation to use the compress option with the ‘KW’ modifier (new to SAS version 9) to only keep writable characters.

The paper includes several worked examples as well as useful code fragments.

The paper was presented at the PharmaSUG conference in Orlando 2010. Download the paper here.

The post SAS paper review #4 appeared first on Statskom.

The paper explores how the FEEDBACK option can be used in the SQL statement  to cause SAS to create additionally detailed log entries which can be useful for debugging. It explains  how use of the FEEDBACK option will cause SAS to write expanded, more explicit messages in the log, which for example explicitly list any variables that were implicitly selected with a SELECT * construction.

We particularly liked the Code Generation section in which Borowiak’s suggests using the FEEDBACK option in combination with the VALIDATE statement to write an expanded version of the SELECT * statement into the log, which can then be copied back into the program and executed.

We feel the paper is suitable for all SAS users who utilise, or are learning SQL.

The paper was presented at the SouthEast SAS® Users Group, 2012 Conference in Raleigh-Durham, North Carolina. Download the paper from the SESUG website.

The post SAS paper review #3 appeared first on Statskom.

The paper explains how Proc Format can be used to combine information from multiple datasets in place  of the traditional “sort/sort/merge” approach and shows how this approach can offer reductions in CPU time of >70% when dealing with large files. In the pharma world this approach has many possible usages. Consider for example the need to look up the RFSTDTC variable in SDTM.DM and merge it onto each of your SDTM datasets in ordering to derive study day variables. Using the author’s approach would instead involve building a format from SDTM.DM which maps USUBJID to RFSTDTC. RFSTDTC can then be attached within the DATAstep using a PUT statement.  

We feel that this approach is particularly relevant now that the widespread adoption of the ADaM model means that companies are having to deal with ever larger analysis datasets. A nice feature of the paper is its proposal for merging with more than one variable.

The paper was presented in the Coder’s Corner section of the SAS Global Forum  2012 Conference in Orlando. Download the paper from Lex Jansen.

The post SAS paper review #2 appeared first on Statskom.

This paper investigates a scenarios that occurs regularly in clinical trials when generating summary tables for categorical data; namely the need to display all possible categories for a CRF even when some categories are not present in the data.

The authors look at what we think is the most commonly used approaches to this problem, i.e. using PROC FREQ/SQL to count the data and forcing out the missing categories by merging the output with a dummy dataset. They then suggest an elegant alternative approach using PROC MEANS in combination with the COMPLETETYPES and PRELOADFMT options.

The paper also suggests how a similar approach can be used to derive total columns.

The paper was presented in the Coder’s Corner section of the PharmaSUG 2012 Conference in San Francisco.

It is available for download here, alternative link.

The post SAS paper review appeared first on Statskom.

The papers we review will all be publically available and will usually have been presented at conferences by organisations such as PhUSE, SAS Global Forum, or PharmaSUG .

Read our first recommendation tomorrow.

The post Introducing our SAS ® paper reviews appeared first on Statskom.