IWI ATE Explorer • DiD estimates: funders × sectors × areas

Overall (pooled)

Metric

ATE (pp) Cost-adjusted (pp per $1M)

Model

Fixed Random (DL)

Cap extreme weights (quantile):

Random uses DerSimonian–Laird. When the metric lacks SEs (cost-adjusted), pooling uses the mean with t-based CI; weight capping has no effect.

Pooled value

95% CI

z / p-value | N

Forest by Funder (pooled across S & U)

Limit to funders:

Min. estimates per funder:

Sort by:

Data snapshot

Rows: 557 | Funders: 22 | Sectors: 45 | Areas: 1

Controls

Funder (F):

Sector (S):

Area (U):

Pooling model

Fixed Random (DL)

Group by:

Min. estimates per group:

Cap weights (quantile):

Sort by:

Download grouped (CSV)

Radar comparison (normalized)

Radar across which dimension?

Min. categories:

Raw data

Notes

Pooled ATE uses inverse-variance weights; Random (DL) adds between-study variance.
When cost-adjusted metrics lack SEs, pooling uses an unweighted mean with t-based confidence interval.
Weight capping reduces oversized influence from extremely small SEs (not applicable when SEs are unavailable).