ImpossibleMA

k=1 - single-study borrowing

Built-in dataset:

Target trial Estimate (log scale): SE: Endpoint:

Adjacent trials (>=3)

Study	Estimate	SE

missing_se - SE reconstruction

Built-in dataset:

Inputs Effect: p-value: CI lower: CI upper: Statistic: df:

Route D - Figure extraction

Upload a horizontal forest plot (PNG or JPG). Click two x-axis reference points and type their values to calibrate. Then click once per study row; drag the proposed whisker-cap handles if needed; confirm to add the row to the extraction table.

Figure: Scale: Confidence:

Calibration (2 x-axis points)

#	pixel x	typed value

Extracted rows

Label	effect	SE	audit

adversarial - worst/best defensible pool

Built-in dataset: or CSV upload:

Studies (>=3 needed; columns: study, estimate, se, rob, n, followup, language, pub_type)

study	estimate	se	rob	n	followup	language	pub_type

ImpossibleMA v0.1

k=1 - single-study borrowing

Envelope

missing_se - SE reconstruction

Envelope

Route D - Figure extraction

adversarial - worst/best defensible pool

Envelope