Ask Mallard
Core concept

What your resources change, and what they don't

4 min read

Resources change what is reachable; they never make an invalid method valid for the same research question.

Fit

When it matters, and when it does not

It matters when

  • You need to match recruitment, data access, skills, software and time to a defensible plan.
  • The proposed analysis requires support your team does not yet have.

It matters less when

Before you start

What to check in your study

Decide these first
  1. People, funding, data access and softwareName who will do each task and what is actually available. Multiple data sources do not by themselves mean record linkage.
  2. The deadline and what must be finishedA proposal deadline is different from a deadline for completed follow-up and results.
It rests on these
  1. The scientific requirement follows from the question and design.How to check: Name the needed support or propose a narrower defensible question. Do not quietly drop clustering, confounding control or missing-data assumptions.
  2. Available records contain the measurements needed for the chosen claim.How to check: Analysis-ready data skip acquisition work only. They do not establish valid measurements, adequate follow-up or causal identification.

Plain language

How it works

Work out a defensible approach first. Then assess access, workload, cost, skills and timing. A pilot can study whether the main study is feasible; it does not establish the main treatment effect.

How it works

Mallard treats Not sure yet as an answer, not a statement that no support or data exist. Unstated funding is not treated as unfunded. Its context checks flag typed conflicts, such as new randomization with already collected outcomes, and ask the investigator to resolve them. Funding and team can raise the drafting route; they cannot lower the route required by study complexity.

The decision in view
Choosing a design from the question. Describing points to case series, prevalence surveys and descriptive cohorts; explaining a cause to randomized trials or observational and quasi-experimental designs under stated assumptions; evaluating a test to accuracy and impact studies; predicting to prediction model development and validation; understanding experience to qualitative and mixed methods; summarizing to systematic and scoping reviews. For a causal question that cannot be randomized, how the exposure arrived points to regression discontinuity, interrupted time series, controlled before-and-after or difference-in-differences, or a cohort emulating a target trial.Start with the kind of questionDescribingCase series, prevalence surveys,descriptive cohortsExplaining a causeRandomized trials; observational andquasi-experimental designs under explicitassumptionsEvaluating a testAccuracy studies; impact on decisions andoutcomesPredictingPrediction model development andvalidationUnderstanding experience or processQualitative and mixed-methods designsSummarizing what is knownSystematic and scoping reviewsFor a causal question, follow how the exposure arrivedCan it be randomized, including clusters or rollout order?Yes: a randomized trialNo: how was it assigned?A threshold rule:regressiondiscontinuityA change on aknown date,measuredrepeatedly:interrupted timeseriesA change in somesites, notothers:controlledbefore-and-afterordifference-in-differencesA treatmentdecision inroutine data: acohort emulatinga target trialEach non-randomized route replaces randomization with an assumption, and that assumption has to bedefensible for this question.
Conceptual diagram. It shows the structure of the idea, not data from any study.

Worked example

A worked example

Invented for illustration
Setting
A clinic has permission to review 100 records before a deadline.
Question
Does this guarantee enough information for an effect study?
Approach
An audit finds that 60 records contain the required outcome.
Result
Only 60 of 100 records, or 60 in 100, have that outcome recorded.
What it means
This describes completeness. The team must assess missingness and attainable precision, obtain more support or change the question before claiming an effect.

In practice

Recent studies PubMed search · not vetted

Searching PubMed…

Watch out

Common mistakes

The one to remember

Context governs feasibility, never validity. Limited time, funding or software cannot make an analysis adequate for a claim it cannot support.

  • Using an independent analysis because clustered software is unavailable.Instead: Obtain suitable support or change the design or question. A spreadsheet does not remove dependence.
  • Treating an unstated budget as no funding.Instead: Leave unknown funding unknown. State a level only when the investigator has supplied it.

Inside Mallard

How Mallard uses this

Mallard uses context to assess feasibility and explain support needs. Funding and team can route drafting upward; they do not lower the scientific requirement.

No Mallard code example points to this page.
Atlas node concept.resources_and_validity, the id a plan cue records.

Related Learn pages

Related designs and methods

Evidence

Sources & further reading

  1. Defining Feasibility and Pilot Studies in Preparation for Randomised Controlled Trials: Development of a Conceptual Framework. (opens in a new tab) · PLoS One · doi 10.1371/journal.pone.0150205
  2. Data sharing and reanalysis of randomized controlled trials in leading biomedical journals with a full data sharing policy: survey of studies published in The BMJ and PLOS Medicine. (opens in a new tab) · BMJ · doi 10.1136/bmj.k400
Prepared with AI assistance. Primary methodological and empirical source records were checked by PMID for title, authors, year and DOI on 2026-11-06. These sources support the general methods; tests pin the separate claims about Mallard. No independent professional approval is claimed.
Last source review: 2026-11-06
AI-generated educational content
Mallard uses reputable methodological and reporting sources, but this material may contain errors or omit important context. Verify important methodological decisions with the cited sources and, when appropriate, a qualified research methodologist or statistician.