VAULT — EDITING, WITH BACKLINKS

Three panes, file tree on the left, backlinks on the right, Markdown in the middle. Nothing here is reinvented. The only differences are in what can appear in a link: questions and source stubs are first-class objects in the tree, in autocomplete, and in the backlinks panel, alongside ordinary notes.

Dissociation designs — power · consolidation-vault
⌘K
⌂
?
▤
◈
⌖
◻
{{ t.label }}
{{ n.glyph }} {{ n.label }} {{ n.tail }}
◆ question◈ source◇ stub
notes / methods / Dissociation designs — power.md edited 3d ago · 412 words
#methods/power #sleep/consolidation #writing/aim2
Dissociation designs — power
The designs that could separate consolidation from encoding strength are all underpowered, and nobody in this literature says so in an abstract. I ran the calculation for the standard nap protocol myself: 0.31 for the effect size I care about.
What the calculation assumed
Two groups of twenty, d = 0.35, α = .05 two-tailed. That is generous — most published nap studies run smaller. See [[◆ Is the overnight benefit consolidation or encoding?]] for where this bears on the argument, and [[◈ Cordi & Rasch 2021#h·12]] for the meta-analytic version of the same point.
The obvious objection is that within-subject designs buy back most of the power. They do, but they also reintroduce the confound the design was meant to remove — which is the substance of [[◆ Is TMR really orthogonal to encoding?]].
Still to check
☐Whether [[◇ Diekelmann 2010]] reports per-group n anywhere — stub, no PDF yet
☑Recompute with the profile-likelihood interval — see [[⌗ run-2026-09-14 · prereg-exclusions]]
“…the available dissociation designs are reviewed and their statistical power is assessed…” — [[◈ Klinzing 2019#h·8]]
BACKLINKS 9 linked · 4 unlinked
{{ b.glyph }} {{ b.title }}
{{ b.context }}
{{ b.meta }}
UNLINKED MENTIONS 4
Lab meeting 04 Sep — “…the power problem in the dissociation designs…”
Grant aim 2 — draft — “…dissociation designs are underpowered…”

TAG BROWSER — SLASH-SEPARATED, PARSED INTO A TREE

The one change to a familiar control: #ml/interpretability/probing is a path, not a string. Selecting a parent includes everything beneath it, and the count on a parent row is the inclusive count — the exclusive count is shown beside it so the difference is never a guess.

TAGS {{ tagSummary }}
{{ t.caret }} {{ t.label }} {{ t.n }} {{ t.own }}
left number — inclusive of children · right number — items tagged at this level only
{{ selTag }} {{ selTagNote }}
{{ r.glyph }} {{ r.title }} {{ r.tag }} {{ r.age }}
Renaming a parent renames the path in every child. Dragging a tag onto another re-parents the whole subtree and rewrites the strings in the files — the tree is a view of the text, never a second source of truth.

SOURCE STUB versus SOURCE WITH A FILE

Both are sources and both are linkable. The distinction is carried by one glyph, a dashed outline, and the absence of the page apparatus — so it is legible in the tree, in autocomplete, in backlinks, and on the page itself. Attaching a PDF is a drop target and a single command, and everything already linked to the stub keeps pointing at it.

◇ SOURCE STUB · NO FILE accepted from scout tmr-watch · 2 Aug
Diekelmann & Born (2010) — The memory function of sleep
venue Nature Reviews Neuroscience 11 doi 10.1038/nrn2762 cited by 3 notes · 1 question
no PDF attached
drop a PDF here attach… ⌘⇧P open DOI ↗
attaching keeps the record and every existing link — the stub becomes the source, it is not replaced
BACKLINKS
Dissociation designs — power — “whether it reports per-group n anywhere”
◆ Is the overnight benefit consolidation or encoding? — supporting, unsorted
No page count, no highlights section, no reader button. The page is short because there is nothing there yet, not because something failed to load.
◈ SOURCE · PDF IN VAULT 14 pages · 4.1 MB · read 47 min
Ólafsdóttir, Bush & Barry (2018) — The role of hippocampal replay in memory and planning
venue Current Biology 28 doi 10.1016/j.cub.2017.10.073 cited by 4 notes · 9 questions
open in reader 37 highlights 12 margin notes
BACKLINKS BY ANNOTATION 6 of 37 highlights are linked
{{ a.id }} {{ a.page }} {{ a.count }}
{{ a.quote }}
{{ l.glyph }} {{ l.title }}
File-level backlinks are listed underneath, as usual. The annotation grouping is the addition — it answers “which of my notes depend on this specific passage”, which file-level backlinks cannot.

GRAPH — PRESENT, NOT PROMOTED

It exists because it is expected, it is honest about what it shows, and it is not the front door. The relational view that earns its place is the Question Map, which is a different surface. Filters are the usual ones plus the two object kinds this vault adds.

Graph 318 nodes · 742 links · showing 2 hops from the current note notes questions sources stubs tags
note question source stub
the Question Map is the view that actually answers questions →