What are some best practices for collaborating on data analytics?

The company and stakeholders goals should drive your data work. Don't build something simply because you want to, build something that will solve a problem at your organization. Set your goals and expectations before any work is done. No one likes moving goal posts, so pick a goal and stick with it.

In the realm of collaborative data analytics, I play a pivotal role across various domains.My capabilities span data exploration, cleaning, statistical analysis and data visualization, including proficiency in natural language processing for textual data. I leverage machine learning algorithms, statistical tests, and visual representations to provide valuable insights. While I excel in algorithm selection, feature engineering, and code assistance in languages like Python or R, it is crucial to recognize the indispensable role of human expertise. The synergy between machine and human intelligence in a collaborative setting ensures a holistic approach to data analytics, facilitating a deeper understanding and informed decision-making process.

A metodologia ágil é uma aliada valiosa nos projetos de análise de dados, oferecendo flexibilidade e capacidade de adapta??o. Ao adotar abordagens como Scrum ou Kanban, as equipes podem iterar rapidamente sobre os dados, garantindo uma entrega mais rápida de valor aos clientes e uma melhor resposta às mudan?as nos requisitos. Além disso, a comunica??o eficaz entre as partes interessadas é promovida, garantindo que as análises estejam alinhadas com os objetivos do projeto.

In my experience the most successful collaborations on data analytics have been to start with explaining the process to the stakeholders of how the analysis will take shape with the 3 D’s: Discovery, Development, Delivery. The Discovery(1) stage is where the data sources are analyzed. Most of the time there is a pretty decent gap between what is actually in the data vs what the stakeholder thinks is in the data. This gap analysis will GREALY reduce the time for Development(2), where “success” is defined using specific fields from specific data source to build the report of dashboard. Finally you Deliver(3) the end product to the stakeholder for User review, and iterate the stages until you land on a finished product.

Your metrics should be align with the problem statement or questions you are trying to answer. Ideally, these metrics/goals are aligned with your stakeholders before you start working on the projects and you both have the same definition of done (setting up expectations)

Do not let your love for technology or tools blind you! While this may sound odd, as tech-savvy folks, we need to keep aside our love for the latest technology features or tools and evaluate tools from a business demands perspective. And even if your business has no clue about analytics tools, they sure have a better understanding of their business needs. So engage and translate those into your tool selection requirements with hard and soft criteria. Some dimensions to consider are Strategic fit, Performance, Ease of Use, Ease of Migration, Adoption and training, Community support, Roadmap, etc.

Most business facing analysis will have the tools and platforms dictated by standards. For example, if we are using a Databricks Lakehouse on Azure with PowerBI being the primary visualization tool, someone would need to business justify why a different tool is needed. It happens … often … but the point of the standards is to make it easier for the majority of people.

Occam's razor can be most helpful here. The simplest way to answer a problem is often (not always, but often) the best way. As a general guideline, never use Tableau when PowerPoint will work, never use a machine learning model when a simple regression/time-series will work, and never use PowerPoint when all that's needed is a couple paragraphs in an email with key data points. Use the simplest tool that can answer the key question.

Choose data analytics tools and platforms wisely. Consider factors like data type, analysis complexity, cost, ease of use, and security. Ensure compatibility, accessibility, and support for effective collaboration.

These are my experience in choosing statistical tool for medical research ? User-Friendly: Prefer simplicity? Go for SPSS. ? Large Datasets: For massive data, lean on SAS. ? Open Collaboration: License worries? Opt for open-source: Python or R. ? Advanced Analysis: Delve deep with Stata for sensitivity and specificity.

Data Governance and Standards are really the cornerstone for collaboration since this is how you ensure you are all speaking the same language. I don't care what tools you have, or what the goals are, if you don't have the same language, collaboration isn't possible (think tower of babel)

In my experience of clinical trial data, Data Governance: ? Standardization: Adopt standards like CDISC (Clinical Data Interchange Standards Consortium) to ensure data consistency across trials. ? Roles & Responsibilities: Clearly define roles such as Data Manager, Data Steward, and Data Custodian to manage, oversee, and protect data. ? Audit Trails: Implement systems that track changes to the dataset, ensuring accountability and transparency. ? Data Security: Ensure stringent measures to protect data from breaches and unauthorized access.

Establishing data governance and quality standards is paramount for reliable data analysis. Define data sources, ownership, and permissions, and implement robust cleaning and validation processes. Document metadata and lineage to ensure transparency and accountability. Regularly monitor and address data quality issues to maintain accuracy and consistency.

In clinical trials, where stakes are high and decisions impact patient lives, the manner in which data analysis is documented and communicated is paramount. Proper documentation ensures that analytical methods, results, and interpretations are transparent, reproducible, and open to scrutiny. It acts as a record of the scientific rigor applied throughout the study.

Solicit feedback regularly to validate data, methods, and results in data analytics. It helps identify errors, gaps, and new insights while building trust with collaborators. Use surveys, comments, or version control to incorporate feedback effectively. Acknowledge and act on received feedback promptly for continuous improvement.

For data analysis of cell therapy data, I find it important to have a cross functional team with people from both the clinical department as well as the CMC (analytics/ process) department. Cell therapy is a living product, derived from a person. The history of the person is influencing the starting material, which in turn impacts the final product. The clinical efficacy depends on the interaction between the final product and the in vitro environment in the patient. Therefore, both the clinical team for the clinical data and the CMC team (analytics/process) for the product data should work with the statisticians, since the specific knowledge is needed for best input, interpretation, and learnings.

Consider the cultural aspects of collaboration in data analytics. Recognize diverse perspectives and foster an inclusive environment where all voices are heard and valued. Embrace curiosity and innovation to drive continuous improvement and adaptation in your analytics endeavors.