Replies: 8 comments 13 replies
-
December 5, 2025
|
Beta Was this translation helpful? Give feedback.
-
December 12th 2025
|
Beta Was this translation helpful? Give feedback.
-
|
Beta Was this translation helpful? Give feedback.
-
|
Usernames to add into pilot 6 repo:
|
Beta Was this translation helpful? Give feedback.
-
|
@rpodcast will handle switching over to zoom and getting zoom recordings for meeting series |
Beta Was this translation helpful? Give feedback.
-
January 9th 2026
|
Beta Was this translation helpful? Give feedback.
-
|
@parmsam-pfizer I created website repo here - https://github.com/[RConsortium/-submissions-pilot6-adams-tlfs-website](https://github.com/RConsortium/-submissions-pilot6-adams-tlfs-website) |
Beta Was this translation helpful? Give feedback.
-
|
Pilot 6 project board here. Private by default but Ben added most of us in. |
Beta Was this translation helpful? Give feedback.
-
|
Please review and edit the meeting summary generated by Zoom and click save to publish the summary to participants via their Individual Dashboard and Public Calendar. Quick RecapThe meeting began with technical discussions about joining the Linux Foundation website and agenda planning, with concerns raised about low attendance and website creation status. Dmitrii presented a pre-release version of the VD Dataset Viewer tool that supports various dataset formats and demonstrated its comparison functionality, while Kevin showcased an AI coding agent that generates code based on dataset specifications. The team discussed modernizing their code to use JSON format and MetaTools packages, with plans to abandon XPTs in favor of JSON format, and addressed data version control and access issues to the project folder on GitHub. Next Steps
SummaryLinux Foundation Meeting Technical Challenges: Project Updates and Meeting Planning:Dmitry mentioned that he had not released the new version yet and was working on it. He planned to show a working version but needed to join from another computer to display more data. Eric and Ben discussed the agenda for an upcoming meeting, with Ben suggesting they use Slack to start discussions and Eric agreeing to update the minutes portal with the new agenda. Steven reported working on the highest law case ADLBHY and requested help transferring data to the Pilot 6 area, which Ben agreed to assist with. Data Version Control Discussion:The team discussed data version control and the use of Excel files versus XML for dataset specifications. Ben agreed to copy over the defined.xml and reference atoms to maintain consistency with Pilot 5. Kevin suggested using Excel files for AI tools, as XML can be difficult for AI to read. The team also addressed access issues to the project folder on GitHub, with Ben offering to add new members. Dmitrii mentioned using Metacore to extract information from defined XML, but acknowledged that Excel might be more convenient for current needs. VD Dataset Viewer Pre-Release Demo:Dmitrii presented a pre-release version of the VD Dataset Viewer on GitHub, which supports comparison of datasets in various formats including SAS7bdate, XPT, and JSON. He demonstrated how to use the Compare tab to select base and compare datasets, showing metadata and data differences. Dmitrii expressed interest in adding RDS support based on user feedback and mentioned that a Mac version would be available in the following week. Dataset Comparison and AI Tools:Dmitrii demonstrated a new dataset comparison tool that will be released next week, featuring functionality to compare differences between datasets and copy data in top-delimited format. Ben and the team discussed the importance of documenting differences between original and current data for Pilot 6, with Dmitrii agreeing that DiffDF is better for documentation than screenshots. Kevin presented an AI coding agent that generates R, SAS, and Python code based on dataset specifications, though it currently requires running on the AI's server rather than local environments. The team discussed the need to modernize the code to use MetaTools and dataset JSON packages, with Ben noting that they should abandon XPTs in favor of JSON format. |
Beta Was this translation helpful? Give feedback.
-
January 16th
|
Beta Was this translation helpful? Give feedback.
-
|
Please review and edit the meeting summary generated by Zoom and click save to publish the summary to participants via their Individual Dashboard and Public Calendar. Quick RecapThe team discussed progress on dataset conversion and validation challenges. They explored options for converting XML files to Excel and JSON formats, with Ben planning to investigate using Pinnacle 21 Community to create Excel specifications. Camilla raised questions about LOCF derivation in the CIBC dataset, which Ben clarified was a known issue from previous pilots. The team also discussed data type differences between XPT and JSON files, with Dmitry explaining the technical considerations around decimal vs. double precision. Regarding tabling packages, the group leaned toward using GT Summary together with Cards, following Daniel's presentation approach, though they agreed to further investigate RTF output capabilities. The conversation ended with updates on dataset progress, including Robert offering to take over the PIX dataset work previously handled by Alana. Next Steps
SummaryData Version Control Workflow Discussion:The team discussed data version control and workflows, with Ben mentioning that Eli's demo was absent due to his unavailability. Dmitry raised questions about saving dataset JSON and suggested libraries for TLFs, while Ben shared his experience with VS Code Cloud and expressed interest in exploring workflows further. Ben also mentioned loading SCTM data into main and converting JSON files, noting that some files might not have been converted due to their irrelevance. He sought help from Dmitry to navigate the code spaces environment and expressed a desire to run the ADCL program using this setup. Script Issues and Editor Solutions:Ben and Dmitry discussed issues with running scripts in Codespaces and the use of Visual Studio Code for R development. They explored the possibility of using a new Posit editor and considered the cost implications of cloud solutions. Ben noted that a script he developed was informed by Dmitry's previous work in ADVS.R. They also discussed the potential to replace XML files with Excel files for specifications, though Dmitry expressed uncertainty about the ease of creating the same output from Excel. Ben mentioned that Excel specifications follow a similar pattern to the current XML process. XML to Excel Conversion Discussion:The team discussed converting an XML file to an Excel file for use with Metacore, with Ben planning to import the Define XML into Pinnacle 21 to generate a specification. Phanikumar raised a question about the necessity of this conversion, to which Ben explained that the data sets are outdated and need to be modernized to avoid Pinnacle 21 errors. The team agreed to test merging the PR for the function change and to update the DSVARS to use the Excel file instead of the XML. Data Format and Validation Challenges:The team discussed challenges with data formats and validation tools. Ben expressed concerns about working with Excel due to version control issues and suggested exploring CSV or JSON formats instead. Phanikumar and Ben considered the advantages of CSV for cross-platform compatibility. Dmitry highlighted the need to convert CSV back to XML, emphasizing the use of Pinnacle 21 Community for this purpose. Steven mentioned using an old XML definition from a CDISC Pilot project without checking for gaps in the standards. The team agreed to move forward with their current approach, despite limitations with the Pinnacle 21 Community tool, and considered reverting to XPTs for easier validation in the future. Dataset Validation Format Conversion Discussion:The team discussed challenges with validating dataset JSON files and the potential need to convert them to XPT files for validation purposes. Dmitry suggested creating an XPT file for validation while maintaining dataset JSON as the resulting file. Ben agreed to create an issue to track this process. They also discussed the possibility of converting dataset JSON 1.1 to XPT and back to dataset JSON to verify no data was lost in the conversion. Ben noted that they should continue using the define in their programs until the Excel file situation is resolved. CIBC Dataset Imputation Discussion:Camilla discussed her work on the CIBC dataset and encountered issues with imputing values using the last observation carried forward method. She noted discrepancies between her results and the original code, particularly regarding the exclusion of Visit 2 in the end-of-treatment derivation. Ben and Phanikumar clarified that there was a previous correspondence with the FDA about this method, suggesting it might be a mistake in the original dataset. Camilla agreed to review the wiki findings for more information and continue her work with a better understanding of the requirements for the variable in question. Data Type Inconsistencies in Files:The team discussed data type inconsistencies between XPT and JSON files, particularly regarding numeric values. Camilla and Ben identified that JSON files can handle different data types (integer, double, float, decimal) compared to XPT files, which only use double. Ben advised Camilla to leave the current data type mismatches documented and suggested creating a wiki to track similar findings from future datasets, similar to what was done in Pilot 3 and Pilot 5. Dmitry inquired about JSON data saving options, noting that while there was previously an option to save all numbers as decimal, they had removed it in favor of using double, float, or integer based on defined XML. Dataset Precision and Team Updates:Ben and Dmitry discussed the use of "float as decimals" in their dataset, with Dmitry explaining that using decimals could lead to unwanted precision and potential mismatches. They agreed to investigate further, with Ben taking the action to look into the reasoning behind their current approach and create an example to discuss next week. In other news, Robert offered to take on the PIX dataset if Alana is no longer available, and Phanikumar mentioned he had just started working on his dataset after a period of illness. Steven, who is currently focused on the DMC, expressed that it would take him some time to make progress on the datasets. Tabling Package Selection Discussion:The team discussed tabling packages, with Ben expressing a preference for GT Summary while acknowledging its integration with ARD functionality. Dmitry, seeking to avoid making a wrong decision, will investigate GT Summary and Cards combination as suggested by Daniel's presentation. Kevin raised questions about RTF output, which Ben will address as an issue, though he personally favors PDFs. The team agreed to focus on selecting a tabling package first, with Ben leaning towards GT Summary, and to revisit the RTF discussion later. ARD Generation and Dataset Limitations:The team discussed ARD (Adverse Reaction Dataset) generation capabilities, with Ben confirming that the cards package can create ARDs, though the FDA team driving the ARS effort has been impacted by recent reorganization changes. Phanikumar inquired about GT summary's ability to generate ARDs, and while Ben acknowledged this was worth investigating, the group agreed to focus on single dataset generation for a single table rather than attempting to feed multiple tables due to current limitations. Dmitry agreed to examine the cards and GT summary packages, with plans to recreate a process using a specific dataset to test ARD capabilities, as ARDs aim to standardize dataset representation. |
Beta Was this translation helpful? Give feedback.
-
January 21st
|
Beta Was this translation helpful? Give feedback.
-
|
Please review and edit the meeting summary generated by Zoom and click save to publish the summary to participants via their Individual Dashboard and Public Calendar. Quick RecapThe team met to discuss data set development and file management processes. They reviewed recent changes to file paths and decided to remove trailing slashes from file paths. Camila shared updates on her work with datasets and received clarification on using Excel files for specifications. The team discussed using Google Sheets as an alternative to Excel for version control of specifications. Eli presented a pull request for implementing DVC (Data Version Control) for managing large datasets, which would help address GitHub storage limitations. The team agreed to investigate DVC further and determine appropriate storage solutions for both specifications and datasets. They set a goal to complete all datasets by the end of February and discussed the need to review helper functions used in their code. Next Steps
SummaryData Review and Documentation Updates:The team reviewed last week's actions and discussed data sets. Ben moved a file to the Atom reference folder as suggested by Dmitry. Camilla looked at the old wiki for OCF derivation and found a different issue in the dataset. She opened a PR with an example to address the AVAL variable problem. The team agreed to document the changes in the wiki, and Camilla offered to investigate the LOCF issue further. Resolving File Path Issues:Camilla discussed issues with file paths in the code, particularly regarding extra slashes that needed to be removed for the code to function correctly on her Mac. Dmitry suggested using global variables for paths to streamline updates, which Camilla acknowledged using but agreed to review further. Ben and Jared noted that file paths had been changing since Pilot 2 or 3, and Sam suggested testing to determine if the extra slashes were necessary for file handling. Camilla agreed to add the slashes back in the pull request for further testing. File Path Slash Removal Discussion:The team discussed removing the slash at the end of file paths in Pilot 5, as Dmitry noticed it was unnecessary and consistent with other systems. Ben confirmed the change worked for him, and Jared suggested testing it further. Camilla agreed to update her PR to remove the slash, and the team aligned on keeping the file path in the admin folder. They also briefly mentioned potentially changing the path to XML in a future PR, but the discussion was cut short due to technical issues. Define File Version Management Issues:The team discussed issues with different versions of the Define file, with Camila encountering errors when using the current repo version compared to a working version she found in the issues. Ben explained that the Define file from Pilot 5 did not include certain datasets like SCIBC, and clarified their process of using Excel files as the source for Define creation, which allows for easier modifications. Dimitri suggested using the most recent CDISC submission package as the source for Excel, which Ben agreed to convert into the Excel format and delete the old version to establish a better source of truth. JSON Data Format Decision Meeting:The team discussed data formats for their project, confirming that JSON will be used for both internal and FDA submissions, while intermediate data will be stored in JSON format for transport and potentially Parquet for internal use. They noted that RDS files are allowed by the FDA, but Parquet files are not. Ben mentioned that they no longer need to use the exporter package since they are not writing XPT files. Camilla and Ben agreed to load data to GitHub for now, with plans to address a better storage solution later. Tool Removal and Data Format Alternatives:The team discussed the removal of several tools, including Library Heaven and Haven, as they are no longer supported. They explored alternatives for outputting SAS 7 VDATs, with Dmitry mentioning the use of XPTs and JSON formats. Jared inquired about maintaining metadata information when transferring data between R and SAS, to which Ben and Phanikumar responded with various file formats that could be used. Ben and Steven briefly discussed progress on ADLBHY, with Steven mentioning he would have more time to work on it in two weeks. Data Integration and Access Planning:The team discussed keeping data and programs together in a branch, with Ben and Dmitry agreeing on copying the program to create a dataset. Brandon offered to set up a meeting with Parul to help her access their system, as she is US-based. Ben suggested a goal of completing data sets by the end of February, which Steven agreed was feasible. They also discussed a file containing helper functions from a previous Pilot, wondering what to do with it and how it interacts with their efficacy data sets and tables. R Script Organization Strategy Discussion:The team discussed organizing helper functions in R scripts, with Steven suggesting sourcing them in the R profile rather than creating a package to align with FDA preferences. Ben and Dimitry debated the merits of separate files versus a single megafile, with Dimitri advocating for individual files for better tracking of updates. They agreed to focus first on analyzing functions used in datasets, with Ben asking Dimitry to review which functions could be removed, while leaving the broader organizational structure open for future discussion as they approach the tables portion of the project. Excel Version Control Alternatives:The team discussed challenges with version controlling Excel files and explored alternatives. Dmitry suggested using a dedicated branch for updating specifications to reduce conflicts, while others proposed using Google Sheets instead of Excel due to better collaboration features and accessibility. Jared offered to investigate packages that could read Excel files from Google Drive, and the team agreed to continue the discussion in Slack. Steven mentioned restrictions on accessing Google-related websites from his work laptop, which could be a potential issue. DVC Implementation for Large Datasets:The team discussed implementing DVC (Data Version Control) as an alternative to GitHub for managing large datasets, particularly the STTM questionnaire dataset which has become too large for GitHub. Eli shared a pull request demonstrating DVC's functionality, which involves tracking data files through immutable hashes rather than storing the files themselves in Git. The team agreed to investigate DVC further, with Eli planning to consult with Terry about suitable storage solutions for the consortium. They also discussed using Google Sheets for dataset updates, with Steven and Jared agreeing to coordinate file sharing between Google Sheets and the repository. The team aims to complete datasets by the end of February, with Ben encouraging open communication about process concerns. |
Beta Was this translation helpful? Give feedback.
-
February 13th, 2026
|
Beta Was this translation helpful? Give feedback.
-
|
Please review and edit the meeting summary generated by Zoom and click save to publish the summary to participants via their Individual Dashboard and Public Calendar. Quick RecapThe meeting focused on dataset management and table generation for the CDIS Pilot project. Ben led discussions on storing datasets using Eli's DVC platform, which allows authenticated access and version control. The team debated approaches for table generation, with Dmitry advocating for deterministic comparisons while Ben preferred using AI tools for efficiency. They agreed to use GT summary and cards package for table generation, with plans to compare outputs against original tables. Brandon offered to share a video demonstrating AI-based table comparison tools. The group also discussed standards for table validation, considering both human and AI-generated outputs. Steven provided an update on ADLBH dataset completion, stating the code was ready for review. The conversation ended with plans to continue discussions on Slack and explore further collaboration on Pilot 7. Next Steps
Dataset Progress and Template Review:The team discussed the agenda for their meeting, including a dataset check-in and a DVC demo. They agreed to add human tabling questions to the agenda, as Dmitry raised concerns about using the correct templates and tables. Ben provided a brief update on the ADLBH dataset, mentioning that he was nearly finished fixing AI-generated errors and expected to complete it by the following week. Camilla was also expected to finish her work on the dataset the next week, while Steven was noted as being busy but reliable in completing his tasks. Code Completion and Dataset Review:The team discussed the completion of code development, with Steven confirming that the code is ready pending review and quality control against the original dataset. Ben suggested tagging the Pilot 6 teams group for review, and mentioned that Robert had set up the GitHub Teams group. The team also discussed the status of datasets, noting that Peru's input was missing, and Ben expressed excitement about having a coded base for all datasets by the end of February, though he acknowledged that updates would likely be needed. Additionally, Ben raised concerns about accessing multiple sheets in Google Sheets for R, suggesting the need for a function to compile data from multiple sheets into a Metacore object. XML Generation Process Discussion:Ben and Dmitry discussed the process of generating defined XML programmatically, considering alternatives to Pinnacle 21. They explored the possibility of using Google Sheets to update the spec, but Ben expressed hesitation about the effort required. Dmitry suggested Metacor as a potential tool, but Ben clarified that it cannot write out a define file. They ultimately decided to continue using the current Excel file method on GitHub, acknowledging potential merge conflicts but noting that the project would be completed in a few weeks. Data Version Control Platform Demo:Eli demonstrated a data version control platform that provides authenticated access to DVC remotes, allowing users to pull data but restrict pushing to maintain data integrity. Ben expressed interest in using this system for future pilots, particularly for storing CDIS pilot data outside GitHub, and discussed plans to integrate it with ADSL.R code. They agreed to collaborate further on implementing this solution and potentially extending it to other programs, with Ben planning to complete programming for datasets by February. Streamlining Table Reproduction Processes:Ben and Dmitry discussed streamlining table reproduction processes for Pilot 6, focusing on establishing standards for outputs like RTFs or PDFs and using tools like GT summary. Brandon offered to create a workflow for comparing AI-generated or human-generated code with original pilot tables, potentially using a bot that could automatically trigger comparisons. Brandon planned to share a video demonstrating the process and continue working on making it exposable for automatic triggering from GitHub. Table Output Validation Strategies:The team discussed approaches for comparing and validating table outputs between the original CDIS Pilot study and their modernized versions. Ben and Dmitry debated the merits of using AI tools versus deterministic methods for comparison, with Dmitry expressing concerns about relying too heavily on AI due to potential errors. They agreed to use GT summary and cards package for table generation, with Ben suggesting that human review of a subset of outputs would be sufficient, while AI tools could handle the bulk of comparisons. Next steps include Dmitry creating a PR for the population program, Ben sharing the source of truth for original tables, and Brandon sending a video about their platform over the weekend. |
Beta Was this translation helpful? Give feedback.
-
February 20th, 2026
|
Beta Was this translation helpful? Give feedback.
-
Quick RecapThe team discussed transitioning away from using the Exporter package by leveraging dataset JSON for metadata and formatting, aiming to simplify processes and reduce redundancy. They explored the use of GT summary and Dockerator for creating tables, deciding to prioritize PDF outputs for consistency and ease of use. Dimitri shared insights on creating ARD files and proposed using a common approach of generating tables in GT structure before converting to ARD, which the team agreed to adopt. The group also touched on the potential for using AI tools to assist with table generation and quality control in the future. Participants were encouraged to experiment with the tools and contribute to the project's progress. Next Steps
SummaryAI Tools for Quality Control:The team discussed several topics including pull requests, data version control, and the use of AI tools for quality control. Ben introduced Jeff Dickinson, a recent addition to the team, and provided an overview of the project's goals and scope. The team agreed to explore using AI tools from Cadji AI to assist with quality control, though it's important to note this is not for FDA submission but could be used in future submissions. They also discussed the possibility of using original tables from a report to create separate PDFs for comparison purposes. Dataset Label and JSON Package:Ben and Dmitry discussed the need for a dataset label and the potential to rely solely on the dataset JSON package for metadata. They questioned the necessity of certain attributes like keys and formats, with Ben suggesting they could be removed if the JSON package handles them. Eli noted that the exporter functions set attributes on objects, and Dmitry confirmed that the label will be taken from the JSON or Excel file at the end. They agreed to investigate whether the dataset JSON package sorts data and to update legacy functions from Pilot 5. Ben planned to finish reviewing the program after the meeting. JSON Dataset Format Transition:Ben and Kevin discussed the use of the dataset JSON R package to read and write JSON files, replacing the use of XPT files. Jared inquired about the process, and Dmitry clarified that they are using dataset JSON to replace XPT format, which is intended to be the future submission format for FDA. Ben explained that all necessary information is already in the spec, so they can eliminate redundant processes. The team agreed on the approach, and Jared confirmed his understanding. Ben mentioned getting stuck on the Analysis 1 flag, noting subtle differences between different implementations. Dataset Labels and PR Updates:The team discussed dataset labels and agreed to use dataset JSON instead of DF labels. Ben and Dmitry reviewed the status of pull requests, with Robert expected to complete his by mid-next week. They decided to remove unnecessary programs from previous pilots and keep only the ADLBC program. The team also discussed RTF formatting for tables, with Dmitry suggesting it might be easier to create RTFs first and then convert to PDF if needed. Data Table Output Strategy Discussion:The team discussed formatting and output approaches for their data tables. They decided to use PDFs instead of RTFs, with Ben recommending the Dockerator package for easy PDF creation. Dimitri will explore using Dockerator to create tables with custom headers and footers. The team agreed to follow the "table to ARD" approach, creating ARDs from GT summary tables rather than creating ARDs first. They discussed organizing their output files, deciding to create separate folders for ARDs and outputs under the tables directory. Ben will add Jeff to the repo and follow up with Eli about DVC later. The team plans to produce datasets and tables from the original pilot and develop a QC process for future submissions. |
Beta Was this translation helpful? Give feedback.
-
@DmitryMK I believe the item in the TODO for you for the |
Beta Was this translation helpful? Give feedback.
-
|
@bms63 I believe that this item was more about helper functions. I have created PR #68 for it, but as I explain in the comment of the PR, I'm not sure we actually need it. We can discuss tomorrow if there is time. |
Beta Was this translation helpful? Give feedback.
-
February 27th
|
Beta Was this translation helpful? Give feedback.
-
|
Please review and edit the meeting summary generated by Zoom and click save to publish the summary to participants via their Individual Dashboard and Public Calendar. Quick RecapThe team met to discuss ongoing work on dataset generation and validation, with a focus on reviewing pull requests and addressing data discrepancies. Camilla and Dmitry worked through issues with LLCF amputation and data imputation, while Steven received guidance on using the save dataset JSON function. The team reviewed Dmitry's table layout work, agreeing to use footnotes in the footer section and maintain formatting consistency. Brandon presented a new double programming workflow for dataset comparison using Trialmind, which includes structural and value-level comparisons with detailed reporting capabilities. The team discussed potential integration of this workflow with GitHub pull requests and agreed to explore using it to identify needed updates to the old CDISC Pilot data. Next Steps
SummaryPull Request Workflow Updates:The team discussed ongoing pull requests and workflow updates. Robert mentioned he would issue a PR either later that day, the following day, or early next week, though it might be larger in columns than needed. Ben proposed reviewing outstanding PRs and any blockers, and Brandon offered to share an initial double programming workflow for Trialmind, which Ben expressed interest in. Camilla noted she had made edits and would provide additional comments either that day or over the weekend. LLCF Data Imputation Process Updates:The team discussed issues with data imputation in the LLCF (Last Observation Carried Forward) process, where Camilla and Dmitry identified discrepancies between their implementation and the CDISC dataset. They agreed that values should be imputed using the Analysis01 flag, which would resolve 10 differences in the dataset. Ben offered to update the Excel spec sheet for Camilla, who lacks access to Excel on her personal computer, and the team aligned on using the DefineXML keys for sorting rather than explicit sorting in the program code. ADLBH Output Updates Review:Ben discussed updates to the ADLBH output and mentioned that Camilla had addressed an issue with decimal values being read as floats. He invited Jeff to review the changes and noted that the dataset would need updates to pass Pinnacle 21 tools due to outdated labels. Ben also inquired about Robert's progress on a PR, suggesting that producing a dataset was the priority rather than achieving a "QComplete" status. Dataset Progress and Technical Updates:Ben discussed the progress of the dataset and mentioned that Robert would issue a pull request by early next week. Dmitry shared his work on a table layout but faced technical difficulties and planned to reconnect with a different computer to share his screen. Phanikumar inquired about the method of generating datasets, to which Ben clarified they were not using AI but discussed using Trialmind for quality control with Brandon. Ben also planned to discuss DVC with Eli for data version control and emphasized the importance of completing these tasks efficiently. Table Design for CSR Document:Dmitry and Ben discussed the design of a table in a CSR document. They agreed that using full-width tables looked better, despite some differences in layout compared to the original CDISC PDF. Ben emphasized that matching numbers was more important than matching margins or headers. They also debated whether to place node information in the footer of the table or in the body, with Dmitry suggesting they should aim for consistency in their approach for all tables. Footnote Placement and R Code Review:The team discussed the placement of footnotes, agreeing to move them to the footer as it's more flexible and aligns with their company's practice. They decided to use the decorator's date function instead of matching the original CSR date. Phanikumar offered to review the R code for sanity checking and work on some tables. Ben expressed hope that this approach could be used as a template for other tables in the future. AI Tools Integration for QA:The team discussed using AI tools like Trialmind and GitHub Copilot to generate tables and reduce manual effort. Dmitry warned about using his code as a template due to potential issues, but the team agreed to refine it further. Brandon presented a structured workflow for QA and double programming, including steps for structural and value-level comparisons, and a final report analyzing discrepancies. Ben praised the report's layout and expressed interest in connecting it to Trialmind for automatic updates through pull requests, which Brandon confirmed was possible but not yet implemented. AI-Driven Data Modernization Project:The team discussed updating old data and using AI tools to modernize it. Brandon explained how to handle discrepancies between old and new data labels, and suggested adding a prompt to ignore expected changes. Phanikumar raised concerns about identifying variables outside of ADAM specifications, which Brandon confirmed the tool could handle. Ben and Eli planned to meet later to discuss the DVC and data hosting on GitHub. The team agreed to continue pushing forward with the project and encouraged others to get involved if they wished. |
Beta Was this translation helpful? Give feedback.
Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
-
Meeting Agenda and Minutes for Pilot 6
Feel free to drop a new agenda or add items or questions!
Beta Was this translation helpful? Give feedback.
All reactions