make stimulus_range a private configuration attribute

psignifit/_configuration.py

@@ -46,12 +46,15 @@ class Configuration:
     pool_max_blocks: int = 25
     priors: Optional[Dict[str, Prior]] = dataclasses.field(default=None, hash=False)
     sigmoid: Union[str, sigmoids.Sigmoid] = 'norm'
-    stimulus_range: Optional[Tuple[float, float]] = None
     thresh_PC: float = 0.5
     verbose: bool = True
     width_alpha: float = 0.05
     width_min: Optional[float] = None
 
+    # private attributes, used for testing and benchmarking -> don't use unless you know
+    # what you are doing
+    _stimulus_range: Optional[Tuple[float, float]] = None
+
     # attributes, if not specified, will be initialized based on others
     bounds: Optional[Dict[str, Tuple[float, float]]] = None
     grid_steps: Dict[str, int] = dataclasses.field(default_factory=dict)
@@ -206,7 +209,7 @@ def check_sigmoid(self, value):
         except KeyError:
             raise PsignifitException('Invalid sigmoid name "{value}", use one of {sigmoids.ALL_SIGMOID_NAMES}')
 
-    def check_stimulus_range(self, value):
+    def check__stimulus_range(self, value):
         if value:
             try:
                 len_ = len(value)

psignifit/psignifit.py

@@ -60,20 +60,19 @@ def psignifit(data: np.ndarray, conf: Optional[Configuration] = None,
 
     levels, ntrials = data[:, 0], data[:, 2]
     if conf.verbose:
-        _warn_common_data_mistakes(levels, ntrials, has_user_stimulus_range=conf.stimulus_range is not None,
-                                   pool_max_blocks=conf.pool_max_blocks)
+        _warn_common_data_mistakes(levels, ntrials, pool_max_blocks=conf.pool_max_blocks)
 
-    stimulus_range = conf.stimulus_range
+    stimulus_range = conf._stimulus_range
     if stimulus_range is None:
         stimulus_range = (levels.min(), levels.max())
 
     width_min = conf.width_min
     if width_min is None:
-        if conf.stimulus_range is None:
+        if conf._stimulus_range is None:
             width_min = np.diff(np.unique(levels)).min()
         else:
             # For user specified stimulus range, use conservative estimate of width_min.
-            width_min = (conf.stimulus_range[1] - conf.stimulus_range[0]) / 100
+            width_min = (conf._stimulus_range[1] - conf._stimulus_range[0]) / 100
 
     bounds = parameter_bounds(min_width=width_min, experiment_type=conf.experiment_type, stimulus_range=stimulus_range,
                               alpha=conf.width_alpha, nafc_choices=conf.experiment_choices)
@@ -127,7 +126,7 @@ def psignifit(data: np.ndarray, conf: Optional[Configuration] = None,
                   data=data)
 
 
-def _warn_common_data_mistakes(levels, ntrials, has_user_stimulus_range, pool_max_blocks) -> None:
+def _warn_common_data_mistakes(levels, ntrials, pool_max_blocks) -> None:
     """ Show warnings for common mistakes.
 
     Checks for too many blocks and too few trials.
@@ -136,7 +135,6 @@ def _warn_common_data_mistakes(levels, ntrials, has_user_stimulus_range, pool_ma
     Args:
         level: Array of stimulus level per block
         ntrial: Array of trial numbers per block
-        has_user_stimulus_range: User configured the stimulus range
         pool_max_blocks: Maximum number of blocks until print of pool warning.
     Returns:
         None
@@ -151,19 +149,15 @@ def _warn_common_data_mistakes(levels, ntrials, has_user_stimulus_range, pool_ma
                       "psignifit.pool_blocks(data).\n"
                       "Hide this warning by increasing conf.pool_max_blocks.")
     # warning if many blocks were measured
-    if len(levels) >= 25 and not has_user_stimulus_range:
+    if len(levels) >= 25:
         warnings.warn(f"""The data you supplied contained {len(levels)}>= 25 stimulus levels.
             Did you sample adaptively?
-            If so please specify a range which contains the whole psychometric function in
-            conf.stimulus_range.
             An appropriate prior prior will be then chosen. For now we use the standard
             heuristic, assuming that the psychometric function is covered by the stimulus
             levels,which is frequently invalid for adaptive procedures!""")
-    if ntrials.max() <= 5 and not has_user_stimulus_range:
+    if ntrials.max() <= 5:
         warnings.warn("""All provided data blocks contain <= 5 trials.
             Did you sample adaptively?
-            If so please specify a range which contains the whole psychometric function in
-            conf.stimulus_range.
             An appropriate prior prior will be then chosen. For now we use the standard
             heuristic, assuming that the psychometric function is covered by the stimulus
             levels, which is frequently invalid for adaptive procedures!""")

psignifit/tests/test_configuration.py

@@ -76,12 +76,12 @@ def test_set_wrong_sigmoid():
 
 def test_set_stimulus_range_wrong_type():
     with pytest.raises(PsignifitException):
-        Configuration(stimulus_range=10)
+        Configuration(_stimulus_range=10)
 
 
 def test_set_stimulus_range_wrong_length():
     with pytest.raises(PsignifitException):
-        Configuration(stimulus_range=(1, 2, 3))
+        Configuration(_stimulus_range=(1, 2, 3))
 
 
 def test_set_width_alpha_wrong_type():
@@ -107,7 +107,6 @@ def test_set_wrong_CI_method():
 
 def test_warning_for_2afc_and_wrong_gamma():
     sigmoid = "norm"
-    stim_range = [0.001, 0.2]
     lambda_ = 0.0232
     gamma = 0.1
 
@@ -116,14 +115,12 @@ def test_warning_for_2afc_and_wrong_gamma():
     options['experiment_type'] = '2AFC'
     options['fixed_parameters'] = {'lambda': lambda_,
                                    'gamma': gamma}
-    options["stimulus_range"] = stim_range
 
     with pytest.warns(UserWarning, match='gamma was fixed'):
         Configuration(**options)
 
 def test_warning_for_equal_asymptote_fixing_lambda_and_gamma():
     sigmoid = "norm"
-    stim_range = [0.001, 0.2]
     lambda_ = 0.2
     gamma = 0.1
 
@@ -132,7 +129,6 @@ def test_warning_for_equal_asymptote_fixing_lambda_and_gamma():
     options['experiment_type'] = 'equal asymptote'
     options['fixed_parameters'] = {'lambda': lambda_,
                                    'gamma': gamma}
-    options["stimulus_range"] = stim_range
 
     with pytest.raises(PsignifitException, match='were fixed to different values'):
         Configuration(**options)

psignifit/tests/test_param_recovery.py

@@ -29,7 +29,6 @@ def test_parameter_recovery_2afc(sigmoid):
     options['sigmoid'] = sigmoid  # choose a cumulative Gauss as the sigmoid
     options['experiment_type'] = '2AFC'
     options['fixed_parameters'] = {'lambda': lambda_}
-    options["stimulus_range"] = stim_range
 
     res = psignifit(data, **options)
 
@@ -64,7 +63,6 @@ def test_parameter_recovery_2afc_eta(random_state, eta):
     options['sigmoid'] = sigmoid  # choose a cumulative Gauss as the sigmoid
     options['experiment_type'] = '2AFC'
     options['fixed_parameters'] = {'lambda': lambda_}
-    options["stimulus_range"] = stim_range
 
     res = psignifit(data, **options)
 
@@ -108,7 +106,6 @@ def test_parameter_recovery_fixed_params(fixed_param):
     options = {}
     options['sigmoid'] = sigmoid  # choose a cumulative Gauss as the sigmoid
     options['experiment_type'] = 'yes/no'
-    options["stimulus_range"] = stim_range
     options['fixed_parameters'] = {}
     # we fix it to a slightly off value, so we can check if stays fixed
     options['fixed_parameters'][fixed_param] = sim_params[fixed_param]
@@ -144,7 +141,6 @@ def test_parameter_recovery_YN(sigmoid):
     options['sigmoid'] = sigmoid  # choose a cumulative Gauss as the sigmoid
     options['experiment_type'] = 'yes/no'
     options['fixed_parameters'] = {'lambda': lambda_}
-    options["stimulus_range"] = stim_range
 
     res = psignifit(data, **options)
 
@@ -177,7 +173,6 @@ def test_parameter_recovery_eq_asymptote(sigmoid):
     options['sigmoid'] = sigmoid  # choose a cumulative Gauss as the sigmoid
     options['experiment_type'] = 'equal asymptote'
     options['fixed_parameters'] = {}
-    options["stimulus_range"] = stim_range
 
     res = psignifit(data, **options)
 

psignifit/tests/test_priors.py

@@ -57,4 +57,4 @@ def func(x):
 
     # For a fixed value, the integral should be one
     norm = psignifit._priors.normalize_prior(func, (1, 1))
-    assert np.allclose([1], norm(1))
+    assert np.allclose([1], norm(1))