Insights · 21 essays

Insights.
Methods and scientific production.

Argument-driven technical analyses, calibrated for researchers who need to publish in indexed journals. In Portuguese and English, published under the Aria Research name.

Network of 35 nodes connected by thin lines representing a field's citation structure in three clusters; one central node highlighted in gold marks a high-centrality author — a canonical reading.
Data and statistics 4 min

Bibliometric analysis as empirical thesis argument

Asserting a gap by subjective reading is fragile under examination. Bibliometrics demonstrates the gap empirically and identifies the authors whose work the manuscript cannot ignore without losing credibility.

bibliometricsliterature reviewLotka's law
AI and machine learning 4 min

LDA vs. BERTopic in academic corpora

LDA models probabilistic mixture over words; BERTopic clusters documents by dense semantic similarity. The choice between the two depends on the evaluative dimension relevant to the analytical objective.

topic modelingBERTopicLDA
AI and machine learning 4 min

Semantic embeddings for systematic review screening

Large-scale manual screening has a 5-12% human error rate and zero documented traceability. Semantic embeddings preserve recall above 90% and make every exclusion auditable against a declared threshold.

systematic reviewsemantic embeddingsSBERT
Data and statistics 4 min

Measurement invariance in translated instruments

Group comparisons require empirical evidence of invariance at four levels. Without it, descriptive statistics hide systematic noise the methodological reviewer identifies in seconds.

measurement invariancepsychometricsCFA
Data and statistics 4 min

Multilevel modeling: when MLM is required and when OLS suffices

ICC below 0.05 allows robust OLS; between 0.05 and 0.20 requires cluster-correction or MLM; above 0.20 MLM is mandatory. The rule methodological reviewers check before the second page.

multilevel modelingMLMICC
Writing and publishing 4 min

Responding to reviewers: defend with data, concede with dignity

Major revision carries an 84.7% final acceptance rate in Q1 medical-scientific journals. The response letter decides whether the manuscript crosses that window or loses in it.

response to reviewerspeer reviewmanuscript submission
AI and machine learning 10 min

Computer vision in medical imaging: high AUC is not enough

Computer vision pipelines for medical imaging fail in Q1 journals not for the accuracy metric but for the absence of documented external validation, demographic subgroup breakdown, and explicit human-in-the-loop intervention. Models with internal AUC of 0.95 drop to 0.54 on data from another hospital, and the STARD-AI, TRIPOD+AI, and CLAIM frameworks consolidated this editorial expectation between 2020 and 2025.

computer visionmedical imagingdeep learning
Data and statistics 11 min

A p-value alone won't cut it: Q1 reviewers read your results section

Q1 journals did not ban the p-value; they banned the p-value standing alone. Reviewers today open a results section looking for four elements in the minimum reporting package post-ASA 2016: effect size, confidence interval, statistical power justification, and substantive interpretation kept distinct from inferential interpretation.

statisticseffect sizep-value
Writing and publishing 7 min

Desk rejection is not an English problem, but a weak contribution

Immediate rejection at a Q1 journal rarely comes down to weak English. In four out of five cases the desk reject is decided by miscalibration between the paper's thesis and the venue's stated mission, by how clearly the abstract delivers the contribution, and by coherence between method and results sections.

peer reviewdesk rejectionacademic writing