Merge pull request #34 from CausalInference/2026-02-devel

Make main estimate and bootstrapping deterministic for a seed

Author: remlapmot

.github/workflows/python-app.yml

@@ -1,5 +1,3 @@
-# This workflow will install Python dependencies, run tests and lint with a single version of Python
-# For more information see: https://docs.github.com/en/actions/automating-builds-and-tests/building-and-testing-python
 name: Test pySEQTarget
 
 on:
@@ -13,39 +11,34 @@ permissions:
 
 jobs:
   test:
-    runs-on: ubuntu-latest
+    runs-on: macos-26
     strategy:
       matrix:
         python-version: ["3.11", "3.12", "3.13", "3.14"]
-    
+
     steps:
     - uses: actions/checkout@v6
-    
+
+    - name: Install uv
+      uses: astral-sh/setup-uv@v7
+
     - name: Set up Python ${{ matrix.python-version }}
-      uses: actions/setup-python@v6
-      with:
-        python-version: ${{ matrix.python-version }}
-    
+      run: uv python install ${{ matrix.python-version }}
+
     - name: Install dependencies
       run: |
-        python -m pip install --upgrade pip
-        pip install flake8 pytest pytest-cov
-        if [ -f requirements.txt ]; then pip install -r requirements.txt; fi
-    
-    - name: Install pySEQTarget package
-      run: |
-        pip install -e .
-    
+        uv venv --python ${{ matrix.python-version }}
+        uv pip install flake8 pytest pytest-cov
+        uv pip install -e .
+
     - name: Lint with flake8
       run: |
-        # stop the build if there are Python syntax errors or undefined names
-        flake8 . --count --select=E9,F63,F7,F82 --show-source --statistics
-        # exit-zero treats all errors as warnings
-        flake8 . --count --exit-zero --max-complexity=10 --max-line-length=127 --statistics
-    
+        uv run flake8 . --exclude=.venv --count --select=E9,F63,F7,F82 --show-source --statistics
+        uv run flake8 . --exclude=.venv --count --exit-zero --max-complexity=10 --max-line-length=127 --statistics
+
     - name: Test with pytest
       run: |
-        pytest tests/ -v --cov=pySEQTarget --cov-report=xml
+        uv run pytest tests/ -v --cov=pySEQTarget --cov-report=xml
 
     - name: Upload coverage reports to Codecov
       uses: codecov/codecov-action@v5

docs/conf.py

@@ -12,14 +12,14 @@
 
 version = importlib.metadata.version("pySEQTarget")
 if not version:
-    version = "0.12.1"
+    version = "0.12.2"
 sys.path.insert(0, os.path.abspath("../"))
 
 project = "pySEQTarget"
 copyright = (
-    f"{date.today().year}, Ryan O'Dea, Alejandro Szmulewicz, Tom Palmer, Miguel Hernan"
+    f"{date.today().year}, Ryan O'Dea, Alejandro Szmulewicz, Tom Palmer, Miguel Hernán"
 )
-author = "Ryan O'Dea, Alejandro Szmulewicz, Tom Palmer, Miguel Hernan"
+author = "Ryan O'Dea, Alejandro Szmulewicz, Tom Palmer, Miguel Hernán"
 release = version
 
 # -- General configuration ---------------------------------------------------

pySEQTarget/SEQuential.py

@@ -183,7 +183,7 @@ def bootstrap(self, **kwargs) -> None:
                 setattr(self, key, value)
             else:
                 raise ValueError(f"Unknown argument: {key}")
-        UIDs = self.DT.select(pl.col(self.id_col)).unique().to_series().to_list()
+        UIDs = self.DT.select(pl.col(self.id_col)).unique().sort(self.id_col).to_series().to_list()
         NIDs = len(UIDs)
 
         self._boot_samples = []

pySEQTarget/analysis/_hazard.py

@@ -24,6 +24,8 @@ def _calculate_hazard(self):
 
 
 def _calculate_hazard_single(self, data, idx=None, val=None):
+    if self.seed is not None:
+        self._rng = np.random.RandomState(self.seed)
     full_log_hr = _hazard_handler(self, data, idx, 0, self._rng)
 
     if full_log_hr is None or np.isnan(full_log_hr):
@@ -33,6 +35,8 @@ def _calculate_hazard_single(self, data, idx=None, val=None):
         boot_log_hrs = []
 
         for boot_idx in range(len(self._boot_samples)):
+            if self.seed is not None:
+                self._rng = np.random.RandomState(self.seed + boot_idx + 1)
             id_counts = self._boot_samples[boot_idx]
 
             boot_data_list = []
@@ -83,6 +87,7 @@ def _hazard_handler(self, data, idx, boot_idx, rng):
         data.select(keep_cols)
         .group_by([self.id_col, "trial"])
         .first()
+        .sort([self.id_col, "trial"])
         .with_columns([pl.lit(list(range(self.followup_max + 1))).alias("followup")])
         .explode("followup")
         .with_columns(

pySEQTarget/analysis/_survival_pred.py

@@ -55,6 +55,7 @@ def _calculate_risk(self, data, idx=None, val=None):
         )
         .group_by("TID")
         .first()
+        .sort("TID")
         .drop(["followup", f"followup{self.indicator_squared}"])
         .with_columns([pl.lit(followup_range).alias("followup")])
         .explode("followup")

pySEQTarget/expansion/_selection.py

@@ -20,6 +20,7 @@ def _random_selection(self):
             == self.treatment_level[0]
         )
         .unique("trialID")
+        .sort("trialID")
         .get_column("trialID")
         .to_list()
     )

pySEQTarget/helpers/_bootstrap.py

@@ -77,6 +77,10 @@ def wrapper(self, *args, **kwargs):
         results = []
         original_DT = self.DT
 
+        seed = getattr(self, "seed", None)
+        if seed is not None:
+            self._rng = np.random.RandomState(seed)
+
         self._current_boot_idx = None
         full = method(self, *args, **kwargs)
         results.append(full)
@@ -127,6 +131,8 @@ def wrapper(self, *args, **kwargs):
 
                 for i in tqdm(range(nboot), desc="Bootstrapping..."):
                     self._current_boot_idx = i + 1
+                    if seed is not None:
+                        self._rng = np.random.RandomState(seed + i)
                     tmp = self._offloader.load_dataframe(original_DT_ref)
                     self.DT = _prepare_boot_data(self, tmp, i)
                     if self._offloader.enabled:

pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "pySEQTarget"
-version = "0.12.1"
+version = "0.12.2"
 description = "Sequentially Nested Target Trial Emulation"
 readme = "README.md"
 license = {text = "MIT"}
@@ -56,7 +56,7 @@ Repository = "https://github.com/CausalInference/pySEQTarget"
 "Ryan O'Dea (ORCID)" = "https://orcid.org/0009-0000-0103-9546"
 "Alejandro Szmulewicz (ORCID)" = "https://orcid.org/0000-0002-2664-802X"
 "Tom Palmer (ORCID)" = "https://orcid.org/0000-0003-4655-4511"
-"Miguel Hernan (ORCID)" = "https://orcid.org/0000-0003-1619-8456"
+"Miguel Hernán (ORCID)" = "https://orcid.org/0000-0003-1619-8456"
 "University of Bristol (ROR)" = "https://ror.org/0524sp257"
 "Harvard University (ROR)" = "https://ror.org/03vek6s52"
 

tests/test_reproducibility.py

@@ -0,0 +1,109 @@
+import os
+
+import numpy as np
+import pytest
+
+from pySEQTarget import SEQopts, SEQuential
+from pySEQTarget.data import load_data
+
+
+def _make_seq(seed, **extra_opts):
+    data = load_data("SEQdata")
+    return SEQuential(
+        data,
+        id_col="ID",
+        time_col="time",
+        eligible_col="eligible",
+        treatment_col="tx_init",
+        outcome_col="outcome",
+        time_varying_cols=["N", "L", "P"],
+        fixed_cols=["sex"],
+        method="ITT",
+        parameters=SEQopts(seed=seed, **extra_opts),
+    )
+
+
+def test_hazard_reproducible_with_seed():
+    results = []
+    for _ in range(2):
+        s = _make_seq(seed=42, hazard_estimate=True)
+        s.expand()
+        s.fit()
+        s.hazard()
+        results.append(s.hazard_ratio)
+
+    assert results[0]["Hazard ratio"][0] == results[1]["Hazard ratio"][0]
+
+
+def test_hazard_bootstrap_se_reproducible_with_seed():
+    results = []
+    for _ in range(2):
+        s = _make_seq(seed=42, hazard_estimate=True, bootstrap_nboot=3)
+        s.expand()
+        s.bootstrap()
+        s.fit()
+        s.hazard()
+        results.append(s.hazard_ratio)
+
+    assert results[0]["Hazard ratio"][0] == results[1]["Hazard ratio"][0]
+    assert results[0]["LCI"][0] == results[1]["LCI"][0]
+    assert results[0]["UCI"][0] == results[1]["UCI"][0]
+
+
+@pytest.mark.skipif(
+    os.getenv("CI") == "true", reason="Bootstrap reproducibility test hangs in CI"
+)
+def test_hazard_bootstrap_percentile_reproducible_with_seed():
+    results = []
+    for _ in range(2):
+        s = _make_seq(
+            seed=42,
+            hazard_estimate=True,
+            bootstrap_nboot=3,
+            bootstrap_CI_method="percentile",
+        )
+        s.expand()
+        s.bootstrap()
+        s.fit()
+        s.hazard()
+        results.append(s.hazard_ratio)
+
+    assert results[0]["Hazard ratio"][0] == results[1]["Hazard ratio"][0]
+    assert results[0]["LCI"][0] == results[1]["LCI"][0]
+    assert results[0]["UCI"][0] == results[1]["UCI"][0]
+
+
+@pytest.mark.skipif(
+    os.getenv("CI") == "true", reason="Reproducibility test hangs in CI"
+)
+def test_survival_reproducible_with_seed():
+    results = []
+    for _ in range(2):
+        s = _make_seq(seed=42, km_curves=True)
+        s.expand()
+        s.fit()
+        s.survival()
+        results.append(s.km_data)
+
+    np.testing.assert_allclose(
+        results[0]["pred"].to_numpy(), results[1]["pred"].to_numpy(), atol=1e-14
+    )
+
+
+@pytest.mark.skipif(
+    os.getenv("CI") == "true", reason="Bootstrap reproducibility test hangs in CI"
+)
+def test_survival_bootstrap_reproducible_with_seed():
+    results = []
+    for _ in range(2):
+        s = _make_seq(seed=42, km_curves=True, bootstrap_nboot=3)
+        s.expand()
+        s.bootstrap()
+        s.fit()
+        s.survival()
+        results.append(s.km_data)
+
+    for col in ["pred", "SE", "LCI", "UCI"]:
+        np.testing.assert_allclose(
+            results[0][col].to_numpy(), results[1][col].to_numpy(), atol=1e-14
+        )