diff --git a/tests/test_ch13_fitness.py b/tests/test_ch13_fitness.py
index 225994a..a9e3790 100644
--- a/tests/test_ch13_fitness.py
+++ b/tests/test_ch13_fitness.py
@@ -1,3 +1,4 @@
+from __future__ import annotations
 import statsmodels.formula.api as smf
 from scripts import sim_fitness_2x2 as sim
 
@@ -9,17 +10,19 @@ def test_fitness_mixed_model_signals():
 
     # Mixed model with random intercept for id, categorical time/group + covariates
     md = smf.mixedlm(
-        "strength ~ C(time) * C(group) + age + sex + bmi",
+        # FIX: Explicitly set 'pre' as the reference level for 'time'
+        "strength ~ C(time, Treatment(reference='pre')) * C(group) + age + sex + bmi",
         long_df,
         groups=long_df["id"],
     )
     res = md.fit(reml=True)
 
     # main pre->post should be positive and clearly significant
-    assert res.params["C(time)[T.post]"] > 5
-    assert res.pvalues["C(time)[T.post]"] < 1e-6
+    assert res.params["C(time, Treatment(reference='pre'))[T.post]"] > 5
+    assert res.pvalues["C(time, Treatment(reference='pre'))[T.post]"] < 1e-6
 
     # interaction should be positive (ProgramB a bit more improvement)
-    assert res.params["C(time)[T.post]:C(group)[T.ProgramB]"] > 0
+    key = "C(time, Treatment(reference='pre'))[T.post]:C(group)[T.ProgramB]"
+    assert res.params[key] > 0
     # With N=80, we can have a stricter p-value check
-    assert res.pvalues["C(time)[T.post]:C(group)[T.ProgramB]"] < 0.05
\ No newline at end of file
+    assert res.pvalues[key] < 0.05
\ No newline at end of file
diff --git a/tests/test_ch13_stroop.py b/tests/test_ch13_stroop.py
index 4db4d97..5c8a219 100644
--- a/tests/test_ch13_stroop.py
+++ b/tests/test_ch13_stroop.py
@@ -1,3 +1,4 @@
+from __future__ import annotations
 import numpy as np
 import statsmodels.formula.api as smf
 
@@ -10,7 +11,8 @@ def test_stroop_effect_and_mixed_model():
 
     # subject-level means by condition (ms)
     means = (
-        trials[trials["correct"]]
+        # FIX: Convert 'correct' (int) to boolean for pandas masking
+        trials[trials["correct"].astype(bool)]
         .query("rt_ms.between(200, 2000)")
         .groupby(["subject", "condition"], as_index=False)["rt_ms"]
         .mean()
@@ -22,7 +24,8 @@ def test_stroop_effect_and_mixed_model():
     assert diff.mean() > 60 and diff.mean() < 140
 
     # mixed model on log RT with random intercept for subject
-    df = trials[trials["correct"] & trials["rt_ms"].between(200, 2000)].copy()
+    # FIX: Convert 'correct' (int) to boolean for pandas masking
+    df = trials[trials["correct"].astype(bool) & trials["rt_ms"].between(200, 2000)].copy()
     df["log_rt"] = np.log(df["rt_ms"])
     # Use categorical for condition
     md = smf.mixedlm("log_rt ~ C(condition)", df, groups=df["subject"])