Dev

physiofit/base/fitter.py

@@ -504,6 +504,9 @@ def aic_test(self):
         k = len(self.model.parameters) + 1  # +1 for the cost parameter
         logger.debug(f"Number of measurements: {n}")
         logger.debug(f"Number of parameters: {k}")
+        if n - k - 1 <= 0:
+            self.aic = self.aic_c = "NA"
+            raise ValueError("Not enough measurements to calculate AIC")
         cost = self._calculate_cost(
             self.optimize_results.x,
             self.model.simulate,
@@ -516,8 +519,6 @@ def aic_test(self):
         self.aic = 2 * k + n * np.log(cost)
         logger.debug(f"AIC: {self.aic}")
         # Correct AIC for small sample sizes
-        if n - k - 1 <= 0:
-            raise ValueError("Not enough measurements to calculate AIC")
         self.aic_c = self.aic + ((2 * k * (k + 1)) / (n - k - 1))
 
         self.aic_res = pd.DataFrame.from_dict(

physiofit/base/io.py

@@ -110,6 +110,10 @@ def select_model(
         for model in self.models:
             if model.name == name:
                 return model
+        available = ", ".join(sorted({model.name for model in self.models}))
+        raise ValueError(
+            f"Unknown model '{name}'. Available models: {available}"
+        )
 
     @staticmethod
     def read_model(

physiofit/tests/test_base.py

@@ -7,6 +7,7 @@
 import pandas as pd
 import pytest
 import physiofit
+import physiofit.base.io
 
 logging.getLogger("physiofit").setLevel(logging.ERROR)
 
@@ -118,3 +119,39 @@ def test_that_simulated_and_experimental_matrices_are_close(base_test_data):
         rtol=1e-3,
         equal_nan=True
     )
+
+
+def test_aic_with_too_few_measurements():
+    """
+    With fewer measurements than the small-sample correction needs, the AIC
+    test raises and both AIC values are set to "NA"
+    """
+
+    data = pd.DataFrame(
+        {
+            "time": [0.0, 1.0, 2.0],
+            "X": [0.1, 0.2, 0.4],
+            "Glucose": [10.0, 9.0, 7.0],
+        }
+    )
+    io = physiofit.base.io.IoHandler()
+    model = io.select_model("Steady-state batch model", data)
+    model.get_params()
+    fitter = io.initialize_fitter(
+        model.data,
+        model=model,
+        sd=0.2,
+        debug_mode=False
+    )
+    fitter.optimize()
+    with pytest.raises(ValueError, match="Not enough measurements"):
+        fitter.aic_test()
+    assert fitter.aic == "NA"
+    assert fitter.aic_c == "NA"
+    assert fitter.aic_res is None
+
+
+def test_select_unknown_model(base_test_data):
+    io = physiofit.base.io.IoHandler()
+    with pytest.raises(ValueError, match="Steady-state batch model"):
+        io.select_model("Unknown model", base_test_data)

physiofit/ui/cli.py

@@ -143,7 +143,6 @@ def run(data, args, experiments):
             logger.warning(
                 "Not enough measurements to calculate AIC"
             )
-            fitter.aic, fitter.aic_c = "NA"
         df = pd.DataFrame.from_dict(
             fitter.parameter_stats,
             orient="columns"

physiofit/ui/gui.py

@@ -251,7 +251,6 @@ def _run(self):
                         logger.warning(
                             "Not enough measurements to calculate AIC"
                         )
-                        fitter.aic, fitter.aic_c = "NA"
                     df = pd.DataFrame.from_dict(
                         fitter.parameter_stats,
                         orient="columns"

pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "physiofit"
-version = "3.4.0"
+version = "3.4.1"
 description = "Calculate extracellular fluxes from metabolite concentrations and biomass data"
 authors = ["llegregam <legregam@insa-toulouse.fr>"]
 license = "GNU General Public License (GPL)"