Programming Validation

1
Programming Validation

• Presented By
• Deb Goodfellow, ASG Inc., Cary, NC

2
Overview

• Why we need validation
• What is the process
• Validation Tools
• Validation Methods and steps

3
Why do we need validation?

• Need for accuracy
• Quality deliverable
• Errors in programming
• Different interpretations of requirements

4
Validation Process

• For output, check spelling, layout, format, etc.
• For datasets, variables, formats, lengths,
  labels, etc.

5
Useful and necessary tools for validation

• CRF
• Protocol
• SAP
• Output shells
• Specs

6
First steps

• Check the specs, SAP, CRF to make sure they apply
  and do not conflict with each other
• Make sure you have the information required to do
  the validation

7
Visual Check

• Does the output match the shell
• Are there misspellings in headers, footnotes,
  titles
• Is the alignment of columns proper
• Are any of the data being truncated
• Are the data formatted correctly

8
Spot checking

• In some companies, there is no dataset to compare
  for output.
• Listings may have hundreds of pages
• Get clarification on which specific values need
  to be checked
• Parallel program and visually check the results

9
Parallel Programming

• Independence is necessary
• Determine which variables are required either by
  specs or ??
• proc compare the datasets

10
Proc compare results

• All information in the proc compare output is
  important
• Note no unequal values were found. It passes???
• Using ID in proc, we match. It passes?

11
One method for double check

• We could visually look at all the information in
  the proc compare results
• Easy to miss something with all the information
  given
• Very time consuming each time you need to re-run

12
A programming method

• SASHELP library
• Dataset VCOLUMN provides information about each
  dataset
• Do parallel programming applying the correct
  formats, labels, lengths, etc according to specs
• Add a data step to compare

13
Code

• Proc sql noprint
• Create table production as select distinct
• Name, length, label, format, informat
• From sashelp.vcolumn where
• Memname DATAX and libname ADS
• Order by Name, length, label, format, informat
• Create table validation as select distinct
• Name, length, label, format, informat
• From sashelp.vcolumn where
• Memname DATAX and libname WORK
• Order by Name, length, label, format, informat
• Quit

14
Check

• Data check
• merge production(ina) validation(inb)
• by name length label format informat
• if a and not b then ds prod
• else if b and not a then ds val
• else delete
• Run

15
Overview

• Validation is more than doing a simple proc
  compare and getting No unequal values were
  found
• The more we can check, the more assured we can be
  of having a quality deliverable
• Use our programming skills whenever possible to
  save time and mistakes

16
Remember

• There is absolutely no pressure. If you miss
  something, chances are good that the FDA will
  find it!