# Methods: Overview, resources

## Sections

[qualitative-design-issues](https://globalimpact.gitbook.io/untitled/methodological-discussion/qualitative-design-issues "mention"): How to design the 'content' of experiments and surveys to have internal validity and external generalizability

[splits-randomization](https://globalimpact.gitbook.io/untitled/methodological-discussion/splits-randomization "mention"): How to set up trials to have comparable groups

[adaptive-design-sampling-reinforcement-learning](https://globalimpact.gitbook.io/untitled/methodological-discussion/adaptive-design-sampling-reinforcement-learning "mention"): Adjusting the treatments and design as you learn, to 'get to the highest value in the end'

[analysis-statistical-approaches](https://globalimpact.gitbook.io/untitled/methodological-discussion/analysis-statistical-approaches "mention"): How to make inferences from the data after you have it (and plan this in advance)

## What are our estimation goals?

### Statistical power versus optimized learning

### Fixed vs adaptive designs

See [adaptive design notes](https://globalimpact.gitbook.io/untitled/methodological-discussion/adaptive-design-sampling-reinforcement-learning)

## Resources

Rethink Priorities notes (some are works in progress)...

{% embed url="<https://rethinkpriorities.github.io/methodology-statistics-design/introduction-and-overview.html>" %}
RP notes on methods
{% endembed %}

{% embed url="<https://rethinkpriorities.github.io/how-to-visualize>" %}
RP dynamic doc on making graphs etc
{% endembed %}

The [https://declaredesign.org/](https://declaredesign.org) framework and R package seems very helpful. I (David Reinstein) am learning and trying to adapt it.

[Dillon's 'Hemlock'](https://dsbowen.github.io/software/hemlock/)

[Reinstein 'research tools and data' airtable list](https://airtable.com/invite/l?inviteId=inv9DBIxRLVztaO8U\&inviteToken=82dc3cdefbad836030c92c23974629c02f0aeb0e93b773c58bb8c7e17d4fa6d6)