MELODEM aims to enhance the rigor of research on Alzheimer's Disease and other causes of dementia via identifying methods problems, sharing best practices, and prompting methodological innovation. Our 2015 guidelines paper outlined common challenges in dementia research.
MELODEM guideline princip 1: Clarify the question. Is it causal or predictive? ID the population. Define the estimand you want. Explain assumptions (causal & statistical) you need to invoke to link your statistical result to your substantive question. alzheimersanddementia.com/article/S1552-…
If you consider your main exposure a target of intervention (eg will reducing blood pressure reduce AD risk?) it's a causal question. If you just want to know if someone is likely to develop AD, but you don't aim to intervene, probably a prediction question.
Best methods for causal and predictive questions differ, for ex because confounders (ie common causes of exposure and outcome) are a problem in answering causal questions, but typically not for strictly predictive questions.
