Ask Mallard beta
AI-assisted study design and biostatistics planning for clinicians.
Describe your study in plain language. Mallard settles the design first, challenges weak assumptions, then builds the plan step by step, from population and data through analysis and sample size, approvals, writing, and runnable code. It is built for clinicians, medical students, residents and trainees planning research with limited statistical support on hand. Wrong statistics can misdirect a real study, so Mallard asks for missing information rather than guessing, and its own code, with no AI in it, runs a fixed set of statistical checks and decides what may be printed.
How to use Mallard
- Choose your design. A free design brief: the design Mallard recommends and why, what the study would estimate, the conditions the design depends on, each with the record-level check that settles it, and the alternatives worth considering. The brief is reviewed by an Anthropic model and, independently, by an OpenAI model.
- Build your plan. Step by step, in the order a study develops: your patients and your data, the analysis and how many you need, approvals, team and cost, then, when you want them, the write-up and the code.
What a plan contains
- The research question and hypothesis, restated precisely
- The study population, with exclusions to report in a flow diagram
- The study design, and the recommended analysis with the reasoning behind it
- Power and sample size, with the assumptions named and a calculator to re-run them on your own numbers
- An editable causal diagram — change the confounders and the plan rebuilds around your edit
- The conditions the design depends on — each with a check a data programmer can run, the number that decides it, and the design to switch to if it fails
- A drafted methods paragraph, abstract, significance, innovation and limitations
- A background section that looks up real PubMed records for each citation
- A read on which IRB pathway is likely, and what the committee decides rather than the tool
- A data-collection plan and questions worth taking to a statistician
- A budget matched to your funding level
- Runnable R, Python, SAS and Stata code, plus point-and-click steps for SPSS
Free sample-size calculators
These run in the browser and need no account:
More
- Methods and validation — how plans are produced, and how the numbers are checked
- Pricing — free plans each month, a Pro subscription, and institutional access for programs and health systems
- Privacy — what is stored, and what never leaves your browser
- Terms
A planning aid, not a substitute
Mallard drafts a plan for a qualified person to check. Its output is a starting draft: verify it with a biostatistician and get approval from your IRB or research ethics committee before collecting data. Do not enter protected health information — a study plan never needs patient identifiers.
The interactive app needs JavaScript. Please enable it to describe a study and generate a plan; the calculators and the pages linked above work without it.