Your team is divided over statistical models. How do you navigate conflicting opinions to reach a resolution?

To navigate conflicting opinions on statistical models, prioritize a data-driven approach by conducting a comparative analysis of each model's performance using the same dataset. Evaluate metrics such as accuracy, robustness, and predictive power to objectively assess the strengths and limitations of each model. For example, in previous projects, I've used cross-validation and performance metrics to objectively determine the most suitable model, which helped resolve disagreements based on empirical evidence rather than subjective preference. This approach ensures that decisions are grounded in objective data, facilitating a consensus based on the model's proven efficacy in addressing the research objectives.

A statistical model is fit for a study depending on the nature of research objectives, the extent to which researchers want to explore the objectives, and time available for researchers to complete the research. Dialogue must center around these points, and certainly others, to build consensus among the divided teams. The facilitator should stress to each team that it’s not about the simplicity or fanciness of a model that matters, it’s about serving the research purpose, use and requirements.

It is necessary to provide an empowering environment that enable team members to exchange ideas and work enthusiastically toward objectives identified in the organization. Besides, motivating, resourcing, and directing team members to passionately and strategically pursue the statistical model chosen is the focus.

I believe that this type of comparison is based on the leader preference, because he will interpret the feedback from team members instead of learning from them. Like John Ruskin said: "Really great men have a curious feeling that the greatness is not in them, but through them."

Very important because it helps a team members gauge their own abilities and understanding of the obstacles you encounter on a daily basis in order to improve your chosen model. Also, it helps to understand where team members are in terms of ability and what to do to help them develop.

Set up a visiting expert series of workshops where an external statistician or data scientist is invited to spend a day with your team. Their fresh perspective and expertise can help identify potential biases or errors, and enhance the overall quality and accuracy of your analysis.

Because rapid changes in technology had been disruptive to any industry it is necessary to hire experts in artificial intelligence (AI), blockchain, machine learning, and other transformation technologies in order to keep up with actual problems in model selection.

While making a decision the most important thing to think of is the use case. What business problem is this statistical approach/model solving. Using that as our North Star, I try to align the team to consensus on how each of the approaches help us answer the question and the pros/cons for each of them across all dimensions including model metrics, maintainability, practicality, data quality etc. This has typically helped drive consensus and decision making to keep solutions closer to user needs.