%202.png)
Longitudinal and Multi-Study Design
Build a research programme where each study really depends on the last, with contingency plans and longitudinal designs that survive attrition and measurement drift.
Free. Works on Claude, ChatGPT, Gemini or any assistant that accepts a skill file.
What this skill does
The method, encoded.
Multi-study doctorates fail in two ways. In the first, three studies follow one another without depending on one another: study two would have been designed identically whatever study one found, so the sequence bought nothing and the thesis is three studies stapled together. In the second, the dependency is real and unhedged: study two cannot be designed until study one returns something usable, study one returns nothing usable in month eighteen, and the timeline is gone.
This skill builds the sequencing logic and tests whether it is real, using one blunt check: write down what study two would look like under two different outcomes of study one. If it is the same design either way, the sequence is a chronology. It then forces contingency branches for every dependency, including a branch that works under a null result, each with a decision date early enough to execute.
For longitudinal work it covers the threats that only appear when you measure twice. Panel, cohort and repeated cross-section support different claims and are routinely confused. Attrition is never random, so who left matters more than how many. Measurement invariance decides whether observed change is real change or measurement change. Age, period and cohort effects cannot all be separated by any design, which is an identification problem rather than a modelling one.
It also handles the publication-based thesis, where three publishable papers do not automatically make a thesis, and the honest feasibility assessment that prevents the commonest cause of doctoral overrun.
Best used for
- Designing a doctoral programme of linked studies
- Planning what study two becomes under each outcome of study one
- Choosing between panel, cohort and repeated cross-section
- Making a defensible claim about change over time
- Building coherence across a publication-based thesis
- Assessing honestly whether a programme fits the timeline
Typical inputs
What you give it.
The overarching programme-level research question, Time and funding actually available, including fixed deadlines, Access status for each intended data source, Results from any completed study, Ethics approval timelines per phase, Attrition estimates from comparable studies, Instruments intended for repeated administration, Institutional rules for publication-based theses
Typical outputs
What you get back.
Programme question and study map with the cost of deleting each study, Named sequencing logic per link with the dependency demonstrated, Contingency branch table with decision dates and null-result paths, Longitudinal design specification stating what it cannot support, Wave structure justified by the phenomenon's timescale, Attrition retention plan and per-wave attrition analysis specification, Measurement invariance plan and instrument change rule, Age, period and cohort identification statement, Mixed-methods integration point with a pre-committed divergence approach, Reverse schedule, feasibility verdict and ranked reduction plan, Publication-based coherence pack and authorship contribution statements
Method coverage
What the skill works through.
- Making each study earn its place in the programme
- Sequencing logic and how to test whether it is real
- Contingency planning for when study one fails
- Panel, cohort and repeated cross-section: what each supports
- Wave spacing on the phenomenon's timescale
- Attrition, and why it is never random
- Measurement invariance: is the change real or is it the instrument?
- Age, period and cohort confounding
- Specifying the integration point in sequential mixed methods
- Assessing feasibility by scheduling in reverse
- Publication-based theses: three papers do not make a thesis
- Academic integrity and AI use
Download
Free skill. One file.
Enter your email once. Every skill you download after that takes a single click.
How to install
Add the skill file and the five kernel protocols to a Claude Project, a ChatGPT Project, a Gemini Gem, or paste them at the top of any assistant conversation. Then give it your real research material, not a description of it.
Download skillQuestions
Common questions.
How do I know if my multi-study design has a real sequencing logic?
Write down what study two would look like under two different outcomes of study one. If the design is the same either way, the sequence is a chronology rather than a dependency, and the studies could run in parallel, which is usually faster. In a real sequence, study two's sampling, instrument or focus is determined by what study one returned.
What if my first study produces a null result?
Plan for it in advance, because a null is one of the most likely outcomes of a well-designed study of an uncertain effect and it is the one candidates have no plan for. Write a version of study two that works under the null, set a decision date early enough to execute the branch, and include a branch for the case where study one fails to collect adequate data at all.
What is the difference between a panel study and a repeated cross-section?
A panel follows the same units over time and is the only design that supports claims about individual change and trajectories. A repeated cross-section draws a fresh sample each wave, supports population-level change, and cannot say anything about individual movement. A flat population trend across waves is entirely compatible with large individual movement in both directions.
Why does measurement invariance matter in longitudinal research?
Because the same items administered twice do not automatically measure the same thing. Meaning drifts, social acceptability changes, and respondents change how they use a scale. Without invariance evidence, an observed change in means may be measurement change rather than real change. Where formal testing is not possible, keep wording identical, never replace an instrument without carrying the original alongside, and document every change.
Can a longitudinal design establish causation?
It establishes temporal order, which is necessary but not sufficient. Time-varying confounding, reverse causation within the measurement interval, and selection into exposure all survive repeated measurement. Causal claims need an identification strategy, not simply more waves.
How do I make a thesis by publication coherent?
The linking material is the thesis. It has to state the overarching contribution no single paper makes, reconcile construct definitions that drifted between papers, make the dependency between papers visible, explain overlaps, document your own contribution to each co-authored paper specifically, and provide an integrated discussion that no individual paper's discussion could, because each was constrained to its own scope.
The skill chain
Works well with.
Research where people already are.
Analyse it where you already work.
Yazi helps researchers conduct surveys, AI interviews and longitudinal research directly through WhatsApp.
