many truths in ecology example

We know the teams fit many models, it was complex, to derive an analytic conclusion to a domain question.

I was just reflecting on a network meta-analysis I worked on in chronic pain, it would be great to come up with an analogous example from ecology.

many facets of chronic pain

I was not asked to perform one network meta-analysis. Instead I was asked to perform a series of them. More than half a dozen, asking which treatment was most effective for of chronic pain?.

nma required:

• Adverse events
• Amitriptyline review (test dataset)
• Moderate pain relief
• Mood outcomes
• Pain intensity
• Physical function
• Quality of life
• Sleep outcomes
• Substantial pain relief
• Withdrawal outcomes

All of these form a complex tapestry of how effective is a particular treatment for chronic pain. None in isolation is sufficient.

That tapestry should be reported to domain experts to evaluate collaboratively and draw conclusions about recommendations. There was no single number for chronic pain.

many models fit

I will push hard for us emphasising that hypothesis testing is not the only way to perform data analysis (hat tip, Tukey's Exploratory Data Analysis) , but more than that, many hypotheses need to be evaluated to draw an operational conclusion for a domain-level question. There is no true model.

It is, however, fascinating to ask, what are the ways they decomposed this domain question? What did they weigh? To what extent is there consensus?

The experiment answers this for this domain question for this largeish population of analytic teams. I think any methodologist would find that interesting.

My many-NMA for chronic pain is true but too far removed from the domain this applies to for it to land in a punchy way.

Handing over to my domain expert to come up with an analogy. The point is modelling plurality is a feature not a bug.