List every study design this server can plan, check, and audit.

Use this first when you do not yet know which design a study is, or to show a user the options. Each row names the design, an id you pass to the other tools, and the claim that design is able to establish.

Designs covered: lesion network mapping, voxel-based lesion-symptom mapping, DBS sweet-spot/probabilistic stimulation mapping, DBS network mapping (connectivity to the stimulation volume), convergent multi-modal mapping, and coordinate network mapping of published findings.

Returns: a table of id, name, and what the design establishes. Don't use when: you already know the design id — call lnm_outline_paper directly.

Decide which lesion/stimulation mapping design a described study belongs to.

Pass a free-text description of the data: the cohorts you have, where they came from, what the outcome variable is, and whether a second cohort exists for validation. The tool scores every design against that description and returns a ranked list with the signals it matched, plus the analyses reviewers of the winning design expect before they will believe the central claim.

Example input: "We traced 21 published lesion cases causing acute-onset vertigo, compared them against 30 control lesions causing other brainstem syndromes, and seeded connectivity in the GSP1000 connectome."

Returns: best match, a confidence flag, a ranked table, and the mandatory analyses. Error handling: when nothing matches, the response says which facts to add to the description rather than guessing a design.

Produce a section-by-section plan for a paper of a given design.

The plan gives, per section: the subsections in the order the literature uses them, a word budget, what the section has to accomplish, the reporting elements that must appear in it, and sentence frames to start from. Before the skeleton it lists the gaps: analyses this design normally has to defend that nothing in the study you described covers.

Call lnm_classify_design first if you do not know the design id.

Args:

• design: a design id or name, e.g. 'lnm', 'dbs-network', 'convergent'

• topic: the symptom, syndrome, or effect being localized, e.g. 'freezing of gait'

• cohorts: one string per cohort, e.g. ['14 published lesion cases', '36 ET patients, VIM-DBS']

• connectome: the normative connectome, e.g. 'GSP1000 rs-fMRI, n=1000'

• analyses_done: analyses already run, e.g. ['leave-one-out cross-validation']

• target_journal: used only to remind you to check word and figure limits

Returns: gaps, a skeleton table, and per-section obligations.

List the long-form guidance documents bundled with this server.

These are the prose references behind the other tools: how each design works, how the Methods are written, which statistics the genre uses and what they are for, figure conventions, and how the Discussion is argued. Filter by design or topic.

Returns: id, title, what it covers, and the resource URI for each document. Fetch one with lnm_get_guidance.

Fetch one guidance document, or one heading within it.

Documents run to several thousand words. Pass 'section' with part of a heading to pull just that heading and its subheadings; without it you get the whole document, truncated at 25,000 characters with a note saying so.

Args:

• id: a document id from lnm_list_guidance

• section: optional, substring of a heading, e.g. 'specificity' or 'permutation'

Error handling: an unknown id lists the valid ids; a heading that matches nothing lists the headings that document actually has.

Return the reporting checklist for a design as a tickable table.

Items are graded: 'required' means a reviewer of this design will ask for it by name, 'expected' means its absence reads as carelessness, 'optional' means it strengthens the paper. Each item carries the reason it matters, so you can decide rather than comply.

Use while drafting the Methods, and again before submission. For a check against finished text instead of a list to work from, use lnm_audit_manuscript.

Args:

• design: optional design id; omitted returns items common to every design

• section: optional, e.g. 'Methods', 'Results', 'Discussion'

• include_optional: default false

Return the sentence frames this literature uses, grouped by section and purpose.

These are the recurring moves: how a lesion cohort is introduced, how a sensitivity result is stated, how a specificity contrast is phrased, how a cross-validated prediction is reported, how a limitation is conceded. They exist to get the register right, not to be pasted in — a paper built from unedited frames reads like a form.

Args:

• section: optional, e.g. 'Abstract', 'Methods', 'Results', 'Discussion'

Return the limitations a reviewer of this design will raise, filtered to the ones that are actually yours.

Every entry pairs the concession with the counter that published papers use, because a bare concession invites the reviewer to finish the sentence. Entries your study already answers are listed separately as 'retired', so you report them as strengths in the Methods instead of conceding them in the Discussion.

Args:

• design: design id

• study_features: what your study did, one string per feature, e.g. ['prospective cohort', 'lesions traced from 3D MRI', 'out-of-sample validation cohort', 'age-matched connectome']. These both trigger and retire entries.

Returns: an ordered list to write as one Discussion paragraph, plus the retired set.

Search a draft for the reporting elements this design requires, and for phrasings that overstate what the design supports.

Findings come in three grades: blocking (a reviewer will reject or major-revise on it), address before submission, and consider. Each finding names the section, the problem, and the fix. Overstatement findings quote the text that triggered them.

This is a text search, not a review. It tells you whether the manuscript says an analysis was done; it cannot tell you whether the analysis was done well. A clean audit is a floor, not a verdict.

Args:

• manuscript: the full text, or at minimum one complete section (120+ words)

• design: optional design id; supplying it adds that design's specific checks

Returns: counts, then findings grouped by severity.