diff --git a/doc/lmm.rst b/doc/lmm.rst
index de93a000..b6d65e6f 100644
--- a/doc/lmm.rst
+++ b/doc/lmm.rst
@@ -106,7 +106,7 @@ B, C₀, and C₁.
     >>> from glimix_core.lmm import Kron2Sum
     >>>
     >>> random = RandomState(0)
-    >>> n = 5
+    >>> n = 15
     >>> p = 2
     >>> c = 3
     >>> Y = random.randn(n, p)
@@ -118,7 +118,7 @@ B, C₀, and C₁.
     >>> mlmm = Kron2Sum(Y, A, X, G, restricted=False)
     >>> mlmm.fit(verbose=False)
     >>> mlmm.lml()  # doctest: +FLOAT_CMP
-    -6.520026228479136
+    -42.08309530985927
 
 We also provide :class:`.KronFastScanner` for performing an even faster
 inference across several (millions, for example) covariates independently.
@@ -174,21 +174,27 @@ The parameters 𝚩ⱼ, 𝚨ⱼ, and sⱼ are found via maximum likelihood.
 .. doctest::
 
     >>> mscanner = mlmm.get_fast_scanner()
-    >>> A = random.randn(2, 5)
-    >>> X = random.randn(5, 3)
+    >>> A = random.randn(2, n)
+    >>> X = random.randn(n, 3)
     >>> r = mscanner.scan(A, X)
     >>> r["lml"]  # doctest: +FLOAT_CMP
-    83.08864898305367
+    -42.74668875515792
     >>> r["effsizes0"]  # doctest: +FLOAT_CMP
-    array([[ 0.01482133,  0.45189275],
-           [ 0.43706748, -0.71162517],
-           [ 0.52595486, -1.59740035]])
+    array([[ 0.00277822, -0.01476164],
+           [-0.0005451 ,  0.00290053],
+           [-0.00990904,  0.05266315]])
     >>> r["effsizes1"]  # doctest: +FLOAT_CMP
-    array([[ 0.03868156, -0.77199913, -0.09643554, -0.53973775,  1.03149564],
-           [ 0.05780863, -0.24744739, -0.11882984, -0.19331759,  0.74964805],
-           [ 0.01051071, -1.61751886, -0.0654883 , -1.09931899,  1.51034738]])
+    array([[-0.0127383 , -0.03796125,  0.04740337, -0.04064709, -0.03945676,
+             0.00239382, -0.01167387,  0.06761218,  0.04603321, -0.00731968,
+            -0.05142721,  0.03228656, -0.02494051, -0.06615618,  0.03947441],
+           [-0.03729972, -0.05256696,  0.01844337,  0.00221303,  0.01784714,
+             0.07925216,  0.03037916, -0.0247654 ,  0.04081066, -0.0442502 ,
+            -0.01950785,  0.00037021,  0.0548026 , -0.03010318, -0.02419531],
+           [ 0.01722535,  0.0408958 , -0.04265975,  0.03336806,  0.02959425,
+            -0.01610653,  0.00428473, -0.05174828, -0.04550623,  0.01396271,
+             0.04619166, -0.02688382,  0.01095342,  0.06031332, -0.02847803]])
     >>> r["scale"]  # doctest: +FLOAT_CMP
-    5.238689482212067e-11
+    1.0452327204999015
 
 API
 ===
diff --git a/glimix_core/lmm/test/test_kron2sum_scan.py b/glimix_core/lmm/test/test_kron2sum_scan.py
index 6e3ed972..e788d035 100644
--- a/glimix_core/lmm/test/test_kron2sum_scan.py
+++ b/glimix_core/lmm/test/test_kron2sum_scan.py
@@ -11,7 +11,7 @@
 
 def test_lmm_kron_scan():
     random = RandomState(0)
-    n = 5
+    n = 20
     Y = random.randn(n, 3)
     A = random.randn(3, 3)
     A = A @ A.T
@@ -37,7 +37,7 @@ def func(scale):
     F1 = random.randn(n, 4)
 
     r = scan.scan(A1, F1)
-    assert_allclose(r["scale"], 0.3000748879939645, rtol=1e-3)
+    assert_allclose(r["scale"], 0.7365021111700154, rtol=1e-3)
 
     m = kron(A, F) @ vec(r["effsizes0"]) + kron(A1, F1) @ vec(r["effsizes1"])
 
@@ -50,13 +50,13 @@ def func(scale):
     assert_allclose(r["lml"], st.multivariate_normal(m, s * K).logpdf(vec(Y)))
 
     r = scan.scan(empty((3, 0)), F1)
-    assert_allclose(r["lml"], -10.96414417860732, rtol=1e-4)
-    assert_allclose(r["scale"], 0.5999931720566452, rtol=1e-3)
+    assert_allclose(r["lml"], -85.36667704747371, rtol=1e-4)
+    assert_allclose(r["scale"], 0.8999995537936586, rtol=1e-3)
     assert_allclose(
         r["effsizes0"],
         [
-            [1.411082677273241, 0.41436234081257045, -1.5337251391408189],
-            [-0.6753042112998789, -0.20299590400182352, 0.6723874047807074],
+            [0.21489119796865844, 0.6412947101778663, -0.7176143380221816],
+            [0.8866722740598517, -0.18731140321348416, -0.26118052682069],
         ],
         rtol=1e-2,
         atol=1e-2,
@@ -66,7 +66,7 @@ def func(scale):
 
 def test_lmm_kron_scan_with_lmm():
     random = RandomState(0)
-    n = 5
+    n = 15
     Y = random.randn(n, 3)
     A = random.randn(3, 3)
     A = A @ A.T
@@ -105,7 +105,7 @@ def test_lmm_kron_scan_with_lmm():
 
 def test_lmm_kron_scan_unrestricted():
     random = RandomState(0)
-    n = 5
+    n = 15
     Y = random.randn(n, 3)
     A = random.randn(3, 3)
     A = A @ A.T