Vertical slice

Steinmetz Visual Decision

Adapter-backed slice contract for the Steinmetz et al. mouse visual discrimination dataset. The harmonizer reads extracted ALF session directories, preserves bilateral left/right contrast variables, and maps source NoGo responses to canonical withhold choices instead of treating them as missing.

No in-browser analysis cell is bundled for this slice yet. Run the pipeline locally with uv run behavtaskatlas steinmetz-visual-decision-report to generate full artifacts.

Findings extracted

Curves attached to this slice, on shared signed-evidence axes.

Metrics

Trials 10,050
Choice trials 6,745
Withhold trials 3,305
Contrast conditions 16

Slice profile

Operational variables for cross-slice comparison.

Species
mouse
Modality
visual
Stimulus metric
bilateral contrast tuple and signed contrast difference
Evidence type
static
Choice type
free-response
Response modality
wheel
Source data level
processed-trial
Analysis outputs
across-session choice surface, session and subject summaries, withhold rate, and binary movement psychometric comparison columns
Data scope
all 39 extracted Steinmetz ALF behavior sessions for family aggregation
Canonical axis
right contrast minus left contrast, with blank trials as zero

Reproduce locally

Copy this recipe to rebuild the slice's artifacts on your machine at the exact commit deployed here.

Reproduce locally

Run the slice pipeline at this commit

git clone https://github.com/aeronjl/behavtaskatlas.git
cd behavtaskatlas
git checkout ec14ee5
uv sync
uv run behavtaskatlas steinmetz-visual-decision-download
uv run behavtaskatlas steinmetz-visual-decision-harmonize
uv run behavtaskatlas steinmetz-visual-decision-analyze
uv run behavtaskatlas steinmetz-visual-decision-report

No --extra group inferred — see the repository README for the right extras flag if the slice has analysis-only deps. Recipe pinned to commit ec14ee5 so a teammate runs the same code path the deploy used.

Artifacts

Slice has a generated local report; the analysis pipeline ran cleanly and produced trial-level output, plots, and provenance.