Dataset Viewer
Auto-converted to Parquet Duplicate
table_id
string
preset
string
config_id
string
replicate
int64
n
int64
truth_ate
float64
truth_cate
float64
archive_path
string
query_path
string
train_0001
nie_wager_a
train_c0001
1
500
0.410321
0.278158
train/tables/train_0001.csv
train/queries/train_0001.csv
train_0002
nie_wager_a
train_c0001
2
500
0.410321
0.278158
train/tables/train_0002.csv
train/queries/train_0002.csv
train_0003
benkeser_binary
train_c0002
1
500
0.051456
0.041841
train/tables/train_0003.csv
train/queries/train_0003.csv
train_0004
benkeser_binary
train_c0002
2
500
0.051456
0.041841
train/tables/train_0004.csv
train/queries/train_0004.csv
train_0005
nie_wager_a
train_c0003
1
500
0.459434
0.446762
train/tables/train_0005.csv
train/queries/train_0005.csv
train_0006
nie_wager_a
train_c0003
2
500
0.459434
0.446762
train/tables/train_0006.csv
train/queries/train_0006.csv
train_0007
benkeser_binary
train_c0004
1
500
0.040753
0.046714
train/tables/train_0007.csv
train/queries/train_0007.csv
train_0008
benkeser_binary
train_c0004
2
500
0.040753
0.046714
train/tables/train_0008.csv
train/queries/train_0008.csv
train_0009
nie_wager_a
train_c0005
1
500
0.562849
0.539844
train/tables/train_0009.csv
train/queries/train_0009.csv
train_0010
nie_wager_a
train_c0005
2
500
0.562849
0.539844
train/tables/train_0010.csv
train/queries/train_0010.csv
train_0011
benkeser_binary
train_c0006
1
500
0.025701
0.022335
train/tables/train_0011.csv
train/queries/train_0011.csv
train_0012
benkeser_binary
train_c0006
2
500
0.025701
0.022335
train/tables/train_0012.csv
train/queries/train_0012.csv
train_0013
nie_wager_a
train_c0007
1
500
0.57458
0.578944
train/tables/train_0013.csv
train/queries/train_0013.csv
train_0014
nie_wager_a
train_c0007
2
500
0.57458
0.578944
train/tables/train_0014.csv
train/queries/train_0014.csv
train_0015
benkeser_binary
train_c0008
1
500
0.056185
0.052962
train/tables/train_0015.csv
train/queries/train_0015.csv
train_0016
benkeser_binary
train_c0008
2
500
0.056185
0.052962
train/tables/train_0016.csv
train/queries/train_0016.csv
train_0017
nie_wager_a
train_c0009
1
500
0.534917
0.63997
train/tables/train_0017.csv
train/queries/train_0017.csv
train_0018
nie_wager_a
train_c0009
2
500
0.534917
0.63997
train/tables/train_0018.csv
train/queries/train_0018.csv
train_0019
benkeser_binary
train_c0010
1
500
0.054028
0.04701
train/tables/train_0019.csv
train/queries/train_0019.csv
train_0020
benkeser_binary
train_c0010
2
500
0.054028
0.04701
train/tables/train_0020.csv
train/queries/train_0020.csv
train_0021
nie_wager_a
train_c0011
1
500
0.574523
0.226408
train/tables/train_0021.csv
train/queries/train_0021.csv
train_0022
nie_wager_a
train_c0011
2
500
0.574523
0.226408
train/tables/train_0022.csv
train/queries/train_0022.csv
train_0023
benkeser_binary
train_c0012
1
500
0.03992
0.038348
train/tables/train_0023.csv
train/queries/train_0023.csv
train_0024
benkeser_binary
train_c0012
2
500
0.03992
0.038348
train/tables/train_0024.csv
train/queries/train_0024.csv
train_0025
nie_wager_a
train_c0013
1
500
0.604205
0.20249
train/tables/train_0025.csv
train/queries/train_0025.csv
train_0026
nie_wager_a
train_c0013
2
500
0.604205
0.20249
train/tables/train_0026.csv
train/queries/train_0026.csv
train_0027
benkeser_binary
train_c0014
1
500
0.043312
0.052742
train/tables/train_0027.csv
train/queries/train_0027.csv
train_0028
benkeser_binary
train_c0014
2
500
0.043312
0.052742
train/tables/train_0028.csv
train/queries/train_0028.csv
train_0029
nie_wager_a
train_c0015
1
500
0.512269
0.551628
train/tables/train_0029.csv
train/queries/train_0029.csv
train_0030
nie_wager_a
train_c0015
2
500
0.512269
0.551628
train/tables/train_0030.csv
train/queries/train_0030.csv
train_0031
benkeser_binary
train_c0016
1
500
0.023966
0.021521
train/tables/train_0031.csv
train/queries/train_0031.csv
train_0032
benkeser_binary
train_c0016
2
500
0.023966
0.021521
train/tables/train_0032.csv
train/queries/train_0032.csv
train_0033
nie_wager_a
train_c0017
1
500
0.589618
0.604232
train/tables/train_0033.csv
train/queries/train_0033.csv
train_0034
nie_wager_a
train_c0017
2
500
0.589618
0.604232
train/tables/train_0034.csv
train/queries/train_0034.csv
train_0035
benkeser_binary
train_c0018
1
500
0.041063
0.041811
train/tables/train_0035.csv
train/queries/train_0035.csv
train_0036
benkeser_binary
train_c0018
2
500
0.041063
0.041811
train/tables/train_0036.csv
train/queries/train_0036.csv
train_0037
nie_wager_a
train_c0019
1
500
0.429634
0.38945
train/tables/train_0037.csv
train/queries/train_0037.csv
train_0038
nie_wager_a
train_c0019
2
500
0.429634
0.38945
train/tables/train_0038.csv
train/queries/train_0038.csv
train_0039
benkeser_binary
train_c0020
1
500
0.046252
0.034509
train/tables/train_0039.csv
train/queries/train_0039.csv
train_0040
benkeser_binary
train_c0020
2
500
0.046252
0.034509
train/tables/train_0040.csv
train/queries/train_0040.csv
train_0041
nie_wager_a
train_c0021
1
500
0.581725
0.936043
train/tables/train_0041.csv
train/queries/train_0041.csv
train_0042
nie_wager_a
train_c0021
2
500
0.581725
0.936043
train/tables/train_0042.csv
train/queries/train_0042.csv
train_0043
benkeser_binary
train_c0022
1
500
0.035112
0.042263
train/tables/train_0043.csv
train/queries/train_0043.csv
train_0044
benkeser_binary
train_c0022
2
500
0.035112
0.042263
train/tables/train_0044.csv
train/queries/train_0044.csv
train_0045
nie_wager_a
train_c0023
1
500
0.2664
0.385355
train/tables/train_0045.csv
train/queries/train_0045.csv
train_0046
nie_wager_a
train_c0023
2
500
0.2664
0.385355
train/tables/train_0046.csv
train/queries/train_0046.csv
train_0047
benkeser_binary
train_c0024
1
500
0.04635
0.026272
train/tables/train_0047.csv
train/queries/train_0047.csv
train_0048
benkeser_binary
train_c0024
2
500
0.04635
0.026272
train/tables/train_0048.csv
train/queries/train_0048.csv

Statistical Inference Arena · ATE and CATE

Estimate the population average treatment effect (ATE) and the conditional average treatment effect (CATE) at a supplied query point from a simulated study of 500 observations. For each target, return a point estimate and a two-sided 95% interval. The collection supports evaluation of classical estimators and tabular foundation models. Release v1.2.2

What do you want to do? Tables Ground truth Download
Train, validate and test a method with local scoring 48 train, 16 dev, 32 test Public Starter kit
Evaluate a method on the official test 2,000 Maintainers score predictions Test inputs and prediction templates

Target and observations

ATE=E[Y(1)−Y(0)],CATE(x⋆)=E[Y(1)−Y(0)∣X=x⋆]. \mathrm{ATE}=E[Y(1)-Y(0)],\qquad \mathrm{CATE}(x_\star)=E[Y(1)-Y(0)\mid X=x_\star].

ATE averages over the population of a study's generating mechanism. CATE is the conditional mean effect at the supplied query. Inputs are baseline covariates $X$, binary treatment $T$ and outcome $Y$, together with $x_\star$ for CATE. Effects are measured in outcome units; binary outcomes use probability differences. Each study contributes one estimate for each target.

Run your first study comparison

Unzip the starter kit, open its directory and run:

python -m pip install numpy
python baseline.py --tables dev/tables --task ATE --output dev_predictions_ate.csv
python score.py --truth dev/truth_ate.csv --predictions dev_predictions_ate.csv --timings dev_predictions_ate.csv.timings.json
python baseline.py --tables dev/tables --task CATE --output dev_predictions_cate.csv
python score.py --truth dev/truth_cate.csv --predictions dev_predictions_cate.csv --timings dev_predictions_cate.csv.timings.json

The example uses a linear T-learner: fit a separate outcome regression in each treatment arm, then contrast their predictions. ATE averages over the study's covariates; CATE uses the supplied query. Percentile intervals come from 200 whole-row bootstrap refits. Across the 16 development studies, ATE gives bias 0.2270, RMSE 0.2382, coverage 31.25% and mean interval length 0.3234. For CATE, the values are bias 0.1992, RMSE 0.3013, coverage 62.50% and mean interval length 0.7601.

Example predictions, timing files and all six metrics are included in the kit. To compare another estimator, produce the same prediction columns and run the scorer. Choose settings on dev/, then evaluate the public test/ split.

Sampling and files

Public starter. Two designs are adapted: Nie–Wager Setup A has five continuous covariates and a continuous outcome; the Benkeser et al. design has two covariates and a binary outcome. Train, dev and test have disjoint parameter configurations. Each configuration generates two independent studies and shares one CATE query. The kit includes the generator and public ATE/CATE labels. Its starter_catalog lists the table ID, preset, configuration, replicate, sample size, truth_ate, truth_cate, observed CSV path and query path. The dev/ directory appears as validation in the viewer. Mechanisms and sources.

Official test. The 2,000 studies cover 1,000 configurations across 13 source families, with two draws per configuration. Each study has an opaque ID, 500 observations, 2–10 covariates and a continuous or binary outcome. The download contains observed CSVs (tables/), CATE queries (queries/), variable types (official_schema.csv), five treatment-stratified folds for each target (folds/ATE/ and folds/CATE/), and two blank prediction templates. Columns are X1 through Xp, T and Y. The official_inputs viewer shows table IDs, sample sizes, covariate counts and variable types. Generation parameters, replicate pairings and effect labels are excluded from this input download; earlier public releases included the labels.

Submit a method

Fix the estimator and settings before running the official test. When using cross-fitting, follow the target's supplied folds and preserve observation order. Complete predictions_ate_template.csv and predictions_cate_template.csv, each with the 2,000 table IDs and these columns:

table_id,estimate,ci_lower,ci_upper

Use finite numeric values and ensure ci_lower <= ci_upper. Send both completed files to the benchmark maintainer for scoring against population ATE and fixed-query CATE. Bias, RMSE, coverage and mean interval length use equal study weights. Variance is computed within paired configurations; inference time is reported when supplied. Timing includes point and interval computation, including bootstrap, and excludes model loading and warm-up. Report the method, interval procedure, data exposure and hardware alongside predictions.

References

  • Balazadeh et al. CausalPFN: Amortized Causal Effect Estimation via In-Context Learning. Paper · Official checkpoint.
  • Nie and Wager. Quasi-oracle estimation of heterogeneous treatment effects. Paper.
  • Wager and Athey. Estimation and Inference of Heterogeneous Treatment Effects using Random Forests. Paper.
  • Athey, Tibshirani and Wager. Generalized Random Forests. Paper.
  • Ye, Shao, Yi and Zhao. Toward better practice of covariate adjustment in analyzing randomized clinical trials. Paper.
  • Benkeser, Carone, van der Laan and Gilbert. Doubly robust nonparametric inference on the average treatment effect. Paper.
Downloads last month
22

Paper for fuyao11/StatsInference-Arena-ATE-CATE