diff --git a/tests/test_read.py b/tests/test_read.py
index dac0cae..60e0462 100644
--- a/tests/test_read.py
+++ b/tests/test_read.py
@@ -17,7 +17,7 @@
   ],
   indirect=True,
 )
-def test_basic_read(simmed_ms):
+def test_regular_read(simmed_ms):
   """Test for ramp function values produced by simulator"""
   xdt = open_datatree(simmed_ms)
 
@@ -33,3 +33,73 @@ def test_basic_read(simmed_ms):
       nelements = reduce(mul, uvw.shape, 1)
       expected = np.arange(nelements, dtype=np.float64).reshape(uvw.shape)
       assert_array_equal(uvw, expected)
+
+
+ANT1_SUBSET = [0, 0, 1]
+ANT2_SUBSET = [0, 1, 2]
+
+
+def _select_rows(antenna1, antenna2, ant1_subset, ant2_subset):
+  dtype = [("a1", antenna1.dtype), ("a2", antenna2.dtype)]
+  baselines = np.rec.fromarrays([antenna1, antenna2], dtype=dtype)
+  desired = np.rec.fromarrays([ant1_subset, ant2_subset], dtype=dtype)
+  return np.isin(baselines, desired)
+
+
+def _excise_rows(data_dict):
+  _, ant1 = data_dict["ANTENNA1"]
+  _, ant2 = data_dict["ANTENNA2"]
+  index = _select_rows(ant1, ant2, ANT1_SUBSET, ANT2_SUBSET)
+  return {k: (d, v[index]) for k, (d, v) in data_dict.items()}
+
+
+@pytest.mark.parametrize(
+  "simmed_ms",
+  [
+    {
+      "name": "backend.ms",
+      "nantenna": 3,
+      "data_description": [(8, ["XX", "XY", "YX", "YY"]), (4, ["RR", "LL"])],
+      "transform_data": _excise_rows,
+    }
+  ],
+  indirect=True,
+)
+def test_irregular_read(simmed_ms):
+  xdt = open_datatree(simmed_ms)
+
+  for node in xdt.subtree:
+    if "data_description_id" in node.attrs:
+      bl_index = _select_rows(
+        node.baseline_antenna1_name.values,
+        node.baseline_antenna2_name.values,
+        [f"ANTENNA-{i}" for i in ANT1_SUBSET],
+        [f"ANTENNA-{i}" for i in ANT2_SUBSET],
+      )
+
+      vis = node.VISIBILITY.values
+      # Selected baseline elements are as expected
+      nelements = reduce(mul, vis.shape, 1)
+      expected = np.arange(nelements, dtype=np.float32)
+      expected = (expected + expected * 1j).reshape(vis.shape)
+      assert_array_equal(vis[:, bl_index], expected[:, bl_index])
+      # Other baseline elements are nan
+      vis = node.VISIBILITY.values
+      assert np.all(np.isnan((vis[:, ~bl_index])))
+
+      uvw = node.UVW.values
+      # Selected baseline elements are as expected
+      nelements = reduce(mul, uvw.shape, 1)
+      expected = np.arange(nelements, dtype=np.float64).reshape(uvw.shape)
+      assert_array_equal(uvw[:, bl_index], expected[:, bl_index])
+      # Other baseline elements are nan
+      assert np.all(np.isnan((uvw[:, ~bl_index, ...])))
+
+      flag = node.FLAG.values
+      # Selected baseline elements are as expected
+      nelements = reduce(mul, flag.shape, 1)
+      expected = np.where(np.arange(nelements) & 0x1, 0, 1)
+      expected = expected.reshape(flag.shape)
+      assert_array_equal(flag[:, bl_index], expected[:, bl_index])
+      # Other baseline elements are flagged
+      assert np.all(flag[:, ~bl_index, ...] == 1)
diff --git a/xarray_ms/testing/simulator.py b/xarray_ms/testing/simulator.py
index f5f0ca4..8d4f821 100644
--- a/xarray_ms/testing/simulator.py
+++ b/xarray_ms/testing/simulator.py
@@ -2,6 +2,7 @@
 import os
 import tempfile
 import typing
+from collections.abc import Callable
 from typing import (
   Any,
   Dict,
@@ -88,6 +89,7 @@ class PartitionDescriptor:
 
 
 DDIDArgType = List[Tuple[npt.NDArray[np.float64], List[str]]]
+PartitionDataType = Dict[str, Tuple[Tuple[str, ...], npt.NDArray]]
 
 
 class MSStructureSimulator:
@@ -124,6 +126,7 @@ def __init__(
     partition: Tuple[str, ...] = ("PROCESSOR_ID", "FIELD_ID", "DATA_DESC_ID"),
     auto_corrs: bool = True,
     simulate_data: bool = True,
+    transform_data: Callable[[PartitionDataType], PartitionDataType] | None = None,
   ):
     assert ntime >= 1
     assert time_chunks > 0
@@ -178,6 +181,7 @@ def __init__(
     self.simulate_data = simulate_data
     self.partition_names = cbp_names
     self.partition_indices = bcbp_indices
+    self.transform_data = transform_data
     self.model = {
       "data_description": self.data_description,
       "feed_map": self.feeds,
@@ -199,6 +203,8 @@ def simulate_ms(self, output_ms: str) -> None:
 
       for chunk_desc in self.generate_descriptors():
         data_dict = self.data_factory(chunk_desc)
+        if self.transform_data is not None:
+          data_dict = self.transform_data(data_dict)
         (nrow,) = data_dict["TIME"][1].shape
         T.addrows(nrow)
 
@@ -311,9 +317,7 @@ def broadcast_partition_indices(
     return np.stack([a.ravel() for a in np.broadcast_arrays(*bparts)], axis=1)
 
   @staticmethod
-  def data_factory(
-    desc: PartitionDescriptor,
-  ) -> Dict[str, Tuple[Tuple[str, ...], npt.NDArray]]:
+  def data_factory(desc: PartitionDescriptor) -> PartitionDataType:
     """Creates simulated MS data from a partition descriptor"""
     try:
       ddid = desc.DATA_DESC_ID.item()