CSS_2016.md

CSS 2016 Task List

In preparation for CSS 2016, Working Group 5, Standard Analyses, highlight a few tasks that we will initiate or complete during our annual symposium.

Our two main subteams are:

1. White Papers (define analyses), and
2. Standard Scripts (implement analyses)

Standard Analyses (White Papers)

Please see this status of current white papers. We will focus on:

1. Reviewing any proposed updates and revisions to 4 previously published papers + including queries raised while implementing Central Tendency analyses
2. Discussing the 3 papers in active development

Standard Scripts

We have 2 main sources for standard scripts:

1. Scripts from the FDA's JumpStart workflow
2. Implementations of Standard Analyses from PhUSE CS White Papers, above

FDA JumpStart Scripts

We will present recent developments during CSS 2016. If you want to get involved immediately in reviewing and testing these scripts, please contact Hanming Tu (see the PhUSE Wiki for email address).

Standard PhUSE CS Analyses (White Paper implementation)

There is still much to do to deliver the analyses in the White Paper on Central Tendency (WPCT). For full details, see our WPCT Task List. From that list, we highlight a few specific tasks, to illustrate the range of ways that you can contribute your expertise. In increasing complexity:

• Learn about the WPCT package from
  • Our PhUSE wiki page
  • Our GitHub WPCT User & Contributor Guide
• Test a Standard Script
  • R: Using our test data or your own ADaM data, produce Fig 7.1 and Fig 7.2 using the corresponding R script
  • SAS: Using our test data or your own ADaM data, produce Fig 7.1, 7.2, 7.3, 7.6, 7.7 and 7.8 using the corresponding SAS script
  • Give us feedback on usability, documentation, design, etc. Help improve the approach and scripts.
• Enhance a SAS or R script
  • Current R and SAS scripts are coded for 1 data domain, such as labs or vitals signs.
  • Analyse and document domain differences. ADaM expertise is highly valuable.
  • Propose an approach or mechanism for handling all intended domains: labs, vital signs, ECG.
  • Note: We have provided space for documenting these domain differences in the docx Requirements Specifications for each WPTC figure.
• Qualification criteria & mechanisms
  • Outline qualification requirements for each WPCT script. What should we test, to confirm the scripts work as expected? What would convince you that they are accurate and robust?
  • Consider & propose mechanisms for qualification. E.g., how to confirm normal range outliers: manual review; output data sets for automated comparison; etc?
  • References:
    • White Paper on Central Tendency
    • Specifications for the WPCT standard analyses,
• PhUSE CS test data
  • Our test data are limited.
  • In conjunction with Qualification criteria, above, describe test data requirements. What test conditions depend on appropriate data? Describe those data.
  • Create those new test data as extensions of our test data, or as new test data.
• Test the SAS macros behind the standard SAS scripts.
  • That SAS Macro Index page shows what has been done ("qualified") and what remains ("implemented"). Click through to see test plans, script, logs and summary results for each qualified macro.
  • Define and implement the tests for one of the remaining SAS macros
  • We have a simple example of how we implement these pass/fail tests for utility macros.
• Spotfire template for WPCT figures
• RStudio Shiny app, to serve & demonstrate R scripts
• Test automation
  • Industry code should behave as expected for different users, in different environments
  • Design and propose a test-automation solution for the repository and code that we have