Foundation
Baseline reconstruction, controlled synthetic benchmarks, and a prospectively specified observational robustness pilot established the initial methodological foundation.
CompletedIndependent research · Stellar astrophysics
TESS QPP
Reproducibility, methodological robustness, and synthetic-domain selection in short-period QPP analysis with TESS.
An independent research project investigating how methodological choices, observational robustness, numerical behaviour, and synthetic known-truth evaluation shape the detection and interpretation of short-period quasi-periodic pulsations in stellar flares observed by TESS.
Manuscript 1 complete
arXiv preprint prepared
Submission pending · Not peer reviewed
Origin
From reproduction to a broader methodological programme
The project began from a relatively focused question: how reproducible is a published short-period TESS QPP analysis when its effective methodological baseline is reconstructed and tested systematically?
That question developed into a broader programme separating observational reproduction, methodological robustness, numerical behaviour, and synthetic known-truth performance.
Reconstructing the baseline required auditing the effective behaviour of the public AFINO workflow. Controlled synthetic experiments then mapped domain-dependent detectability and numerical behaviour, while observational pilot and catalogue-scale phases tested how classifications respond to prospectively specified changes in temporal window and processing.
These evidence layers are deliberately kept distinct. Observational reference labels are not treated as physical ground truth, and synthetic performance is not automatically transported to the real TESS population.
What can be established when reproduction, observational robustness, numerical stability, and synthetic ground truth are evaluated without collapsing their different truth conditions?
Completed programme
Manuscript 1 evidence is complete
The scientific work supporting the first manuscript is complete and frozen across distinct evidence planes.
Catalogue-scale observational robustness
The pilot architecture was extended to 122 observational references with an explicit baseline-reproduction gate, input-admissibility accounting, seed checks, and conditional period robustness.
Completed & frozenSynthetic ground truth & HELDOUT
A separately developed synthetic programme used prospectively separated DEVELOPMENT and independent single-use HELDOUT evidence to characterize the frozen baseline within its designed domain.
Completed & frozenMain methodological result
Different questions require different evidence
The completed programme shows why reproducibility, robustness, numerical stability, and classification performance should not be treated as interchangeable concepts.
At catalogue scale, many published-QPP references do not reproduce their frozen reference state under the project baseline, so subsequent robustness claims are conditioned on baseline-concordant cases rather than interpreted as observational accuracy.
Within the independent synthetic HELDOUT domain, the untouched AFINO 0.5 rule shows a low-sensitivity, very-high-specificity operating profile. Those known-truth metrics remain synthetic-domain results rather than observational performance estimates.
Research roadmap
Manuscript 1 complete, public preprint next
Foundation
F0Baseline reconstruction & auditCompleted
↓
F1Synthetic benchmarks & numerical behaviourCompleted
↓
F2Pilot observational robustnessCompleted
→
Manuscript 1 evidence
F3ACatalogue-scale observational robustnessCompleted
↓
F3BInjection–recovery & independent HELDOUTCompleted
↓
Manuscript 1Robustness, selection properties & validation architectureCompleted
→
Publication & later work
arXivJournal-neutral public preprintSubmission pending
↓
Peer reviewJournal submissionPending
↓
Later phasesPopulation & physical inferenceFuture work
Current stage
Preparing the public preprint
The scientific work for Manuscript 1 is complete.
An author-approved journal-neutral arXiv preprint has been prepared from the frozen scientific content, including a transparency acknowledgment describing selective use of generative AI for editorial and technical assistance.
The immediate next step is public-preprint submission. No arXiv identifier or journal-submission status is claimed here until those events actually occur.
Open research record
Inspect the project history
The public repository preserves the project structure, frozen scientific phases, decision records, manuscript history, and the current arXiv-neutral preprint workspace.
The transparency-amended pre-submission state is frozen under the Git tag manuscript1-arxiv-neutral-v1.1.
Historical protocols, decisions, and freezes are preserved rather than rewritten to match the current publication state.
Open research repositoryResearch principles
How the project is developed
Reproducibility
Methods, decisions, and scientific boundaries are recorded and frozen at explicit checkpoints.
Separation of evidence
Observational reproduction, observational robustness, numerical behaviour, synthetic ground truth, and physical interpretation remain distinct evidence layers.
Bounded claims
Claims are expanded only when the corresponding evidence supports them, and unresolved limitations remain visible in the public record.
Research status
Manuscript 1 complete, arXiv submission pending
TESS QPP remains an active independent research programme. The first manuscript is complete and its journal-neutral preprint is prepared, but the work has not yet undergone peer review.
Major public updates are added at selected scientific checkpoints rather than after every intermediate analysis step.