Description
experimental-design is an agent skill for planning experiments and studies before any data is collected. It helps choose a design, randomise assignments, block units, and lay out the treatment combinations, so the results can answer the question you care about. The skill is built on three ideas. Randomisation assigns treatments at random so that known and unknown confounders are balanced in expectation, which is what turns a comparison into a causal claim. Replication means independent repetition at the right level, so you can estimate variability, and the skill names pseudoreplication, counting repeated measurements on one unit as independent replicates, as the most common fatal error. Blocking groups similar units, such as by batch, day, site or litter, and randomises within the blocks, which removes known nuisance variation from the noise. The skill then helps you pick among design types and generate the actual randomisation or designed-experiment layout with reproducible scripts. It warns against structural mistakes that make data uninterpretable, and it notes that no analysis can rescue a confounded design afterwards.…
Highlights
- Chooses a design before data collection
- Randomised assignment for causal claims
- Blocking to remove nuisance variation
- Warns about pseudoreplication
- Generates reproducible layouts with scripts
Ratings & Reviews
No ratings yet
Be the first to rate experimental-design.
People also visit
hypothesis-generation
Formulates evidence-bounded scientific questions, candidate hypotheses
experimental-design FAQ
When is it used?
Before data is collected, to decide how units are assigned to conditions and how treatments are combined.
What is pseudoreplication?
Counting repeated measurements on the same unit as if they were independent replicates. The skill calls it the most common fatal error.
Why randomise?
Random assignment balances confounders, known and unknown, in expectation, which is what lets a comparison support a causal claim.
Can analysis fix a bad design?
The skill says no analysis can rescue a confounded or pseudoreplicated design after the fact.
How do I install it?
Run npx skills add https://github.com/k-dense-ai/scientific-agent-skills --skill experimental-design.
Is it checked for security?
Its security audits show passes from Gen Agent Trust Hub, Socket and Snyk.
What are the best experimental-design alternatives?
Popular alternatives to experimental-design include scientific-critical-thinking, gget and adaptyv. See all experimental-design alternatives