SearchLibrium 0.0.151__tar.gz → 0.0.152__tar.gz

{searchlibrium-0.0.151/src/SearchLibrium.egg-info → searchlibrium-0.0.152}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: SearchLibrium
-Version: 0.0.151
+Version: 0.0.152
 Summary: A Python package for econometric models driven by search
 Author: Mario Fernando Taco Morales, Alexander Paz, Prithvi Beeramole, Robert Burdett
 Author-email: Zeke Ahern <z.ahern@qut.edu.au>

{searchlibrium-0.0.151 → searchlibrium-0.0.152}/pyproject.toml

@@ -82,7 +82,7 @@ Documentation = "https://pypi.org/project/SearchLibrium/"
 realpython = "SearchLibrium.__main__:main"
 
 [tool.bumpver]
-current_version = "0.0.151"
+current_version = "0.0.152"
 version_pattern = "MAJOR.MINOR.PATCH"
 commit_message = "[skip ci] Bump version {old_version} -> {new_version}"
 commit = true

{searchlibrium-0.0.151 → searchlibrium-0.0.152}/src/SearchLibrium/constraints_builder.py

@@ -135,6 +135,79 @@ class ConstraintBuilder:
         self._print(f"Mutually exclusive group: {variables}")
         return self
 
+    # ========================
+    # Alternative-Specific Variable Constraints
+    # ========================
+
+    def asvar_alt_spec(self, variable: str, alts: List[str]) -> 'ConstraintBuilder':
+        """
+        Break alternative-specific variables into subsets so they don't all
+        apply uniformly to the model. For each variable, creates
+        alternative-specific dummy columns only for the given *alts*, while
+        keeping the original variable for the remaining alternatives as a
+        generic coefficient. The search will then select among these
+        alt-specific columns as part of the variable search.
+
+        Parameters
+        ----------
+        variable : str
+            Base variable name (e.g. 'TIME', 'COST').
+        alts : list of str
+            Alternatives for which alt-specific coefficients should be
+            created (e.g. ['Car', 'Train']). Other alternatives share a
+            single generic coefficient.
+
+        Returns
+        -------
+        ConstraintBuilder
+            Self for method chaining
+
+        Example
+        -------
+        >>> constraints.asvar_alt_spec('TIME', ['Car', 'Train'])
+        >>> constraints.asvar_alt_spec('COST', ['Bus', 'Car'])
+
+        This creates ``TIME_Car``, ``TIME_Train``, ``COST_Bus``, ``COST_Car``
+        in the active variable pool while keeping ``TIME`` and ``COST`` as
+        generic columns for the remaining alternatives.
+        """
+        if 'asvar_alt_spec' not in self._constraints:
+            self._constraints['asvar_alt_spec'] = {}
+        self._constraints['asvar_alt_spec'][variable] = list(alts)
+        self._print(f"Alt-specific subset: '{variable}' → {alts}")
+        return self
+
+    def incompatible_specs(self, *alt_var_names: str) -> 'ConstraintBuilder':
+        """
+        Declare that certain alternative-specific variable names (e.g.,
+        ``TIME_Car`` and ``COST_Car``) are incompatible and must not appear
+        together in the same model. Each call creates one exclusion group.
+
+        This is applied **after** the alternative-specific dummies are
+        created, so you can declare constraints on the resulting column
+        names.
+
+        Parameters
+        ----------
+        *alt_var_names : str
+            Alt-specific column names forming a mutually exclusive group.
+
+        Returns
+        -------
+        ConstraintBuilder
+            Self for method chaining
+
+        Example
+        -------
+        >>> constraints.incompatible_specs('TIME_Car', 'COST_Car')
+        >>> constraints.incompatible_specs('FARE_Train', 'TIME_Train')
+        """
+        if 'incompatible_specs' not in self._constraints:
+            self._constraints['incompatible_specs'] = []
+        self._constraints['incompatible_specs'].append(list(alt_var_names))
+        self._print(f"Incompatible alt-specific group: {alt_var_names}")
+        return self
+
     def min_behavioral(self, min_count: int, *variables: str) -> 'ConstraintBuilder':
         """
         Require at least ``min_count`` variables from a behavioural pool

{searchlibrium-0.0.151 → searchlibrium-0.0.152}/src/SearchLibrium/search.py

@@ -1703,6 +1703,26 @@ class Search():
     ''' ---------------------------------------------------------- '''
     def remove_redundant_asvars(self, asvars, transasvars, asvarnames):
     # {
+        # Gather variables that are part of an asvar_alt_spec constraint so
+        # that both the generic column AND its alt-specific dummies are kept
+        # together (the constraint requires both forms to coexist).
+        constraints = getattr(self.param, 'pres_spec_constr', None) or {}
+        alt_spec_map = constraints.get('asvar_alt_spec', {})
+        constrained_base = set()
+        constrained_dummies = set()
+        if alt_spec_map:
+            for base_var, alts in alt_spec_map.items():
+                if base_var in asvars:
+                    constrained_base.add(base_var)
+                for alt in alts:
+                    dummy_name = f"{base_var}_{alt}"
+                    if dummy_name in asvars:
+                        constrained_dummies.add(dummy_name)
+
+        # Merge constrained vars -- they always travel together so exclude
+        # them from the either/or logic below.
+        constrained_merged = constrained_base | constrained_dummies
+
         # Filter out elements from asvars that contain any substring present in transasvars.
         redundant_asvars = [var for var in asvars if any(subvar in var for subvar in transasvars)]
         unique_vars = [var for var in asvars if var not in redundant_asvars]
@@ -1710,8 +1730,10 @@ class Search():
         # When transformations are not applied, the redundancy is created
         # if a variable has both generic & alt-spec coeffs
         if len(transasvars) == 0:  # {
-            gen_var_select = [var for var in asvars if var in asvarnames]
-            alspec_final = [var for var in asvars if var not in gen_var_select]
+            gen_var_select = [var for var in asvars
+                              if var in asvarnames and var not in constrained_merged]
+            alspec_final = [var for var in asvars
+                            if var not in asvarnames and var not in constrained_merged]
         # }
         else:
         # {
@@ -1719,12 +1741,16 @@ class Search():
             for var in transasvars:
             # {
                 redun_vars = [item for item in asvars if var in item]
-                gen_var = [var for var in redun_vars if var in asvarnames]
+                gen_var = [v for v in redun_vars
+                           if v in asvarnames and v not in constrained_merged]
                 if gen_var:
                     gen_var_select.append(np.random.choice(gen_var))
-                alspec_redun_vars = [item for item in asvars if var in item and item not in asvarnames]
-                trans_alspec = [item for item in alspec_redun_vars if any(sub_item in item for sub_item in boxc_l)]
-                lin_alspec = [var for var in alspec_redun_vars if var not in trans_alspec]
+                alspec_redun_vars = [item for item in asvars
+                                     if var in item and item not in asvarnames
+                                     and item not in constrained_merged]
+                trans_alspec = [item for item in alspec_redun_vars
+                                if any(sub_item in item for sub_item in boxc_l)]
+                lin_alspec = [v for v in alspec_redun_vars if v not in trans_alspec]
                 choice = np.random.randint(2)  # Chooses a 0 or 1
                 ref = lin_alspec if choice else trans_alspec
                 alspec_final.extend(ref)
@@ -1738,6 +1764,8 @@ class Search():
         else:
             final_asvars = alspec_final
 
+        # Append constrained vars (both generic base and alt-specific dummies)
+        final_asvars.extend(list(constrained_merged))
         final_asvars.extend(unique_vars)  # Extend the list
 
         # Remove duplicates while preserving the order of elements
@@ -1771,6 +1799,7 @@ class Search():
         asvars.extend(self.param.ps_asvars)
         asvars = self.remove_redundant_asvars(asvars, self.param.trans_asvars, self.param.asvarnames)
         asvars = self._apply_mutual_exclusion_filter(asvars)
+        asvars = self._apply_incompatible_specs_filter(asvars)
         return asvars
     # }
 
@@ -1785,6 +1814,7 @@ class Search():
         isvars = [self.param.avail_isvars[i] for i in isvar_select_pos]
         isvars.extend(self.param.ps_isvars)
         isvars = self._apply_mutual_exclusion_filter(isvars)
+        isvars = self._apply_incompatible_specs_filter(isvars)
         return isvars
     # }
 
@@ -2084,6 +2114,9 @@ class Search():
         # ── Mutual-exclusion constraint ──
         self._enforce_mutual_exclusion(solution)
 
+        # ── Incompatible alt-specific specs constraint ──
+        self._enforce_incompatible_specs(solution)
+
         # ── Min-behavioural constraint (soft: at least n from a pool) ──
         self._enforce_min_behavioral(solution)
 
@@ -2176,6 +2209,17 @@ class Search():
             return []
         return self.param.pres_spec_constr.get('mutually_exclusive', [])
 
+    def _get_incompatible_specs_groups(self):
+        """Return the list of incompatible alt-specific variable groups, or [].
+
+        Each group is a list[str] of alt-specific column names (e.g.,
+        ``TIME_Car``, ``COST_Car``); at most one member of each group
+        may appear in a solution.
+        """
+        if not hasattr(self.param, 'pres_spec_constr') or self.param.pres_spec_constr is None:
+            return []
+        return self.param.pres_spec_constr.get('incompatible_specs', [])
+
     def _get_excluded_by_mutual_group(self, already_present: set):
         """Return the set of variables that are blocked because a partner
         from the same mutually-exclusive group is *already_present*.
@@ -2187,6 +2231,37 @@ class Search():
                 excluded.update(gset)
         return excluded
 
+    def _get_excluded_by_incompatible_specs(self, already_present: set):
+        """Return the set of alt-specific column names that are blocked
+        because a partner from the same incompatible-spec group is
+        *already_present*.
+        """
+        excluded = set()
+        for group in self._get_incompatible_specs_groups():
+            gset = set(group)
+            if gset & already_present:
+                excluded.update(gset)
+        return excluded
+
+    def _apply_incompatible_specs_filter(self, varlist):
+        """Given a list of variables, keep at most one per incompatible-spec
+        group.  For each group the first variable encountered is kept;
+        subsequent members are dropped.
+        """
+        groups = self._get_incompatible_specs_groups()
+        if not groups:
+            return varlist
+        blocked = set()
+        for group in groups:
+            seen_first = False
+            for v in varlist:
+                if v in group:
+                    if seen_first:
+                        blocked.add(v)
+                    else:
+                        seen_first = True
+        return [v for v in varlist if v not in blocked]
+
     def _enforce_mutual_exclusion(self, solution):
         """Ensure at most one variable per mutually-exclusive group is present.
 
@@ -2215,6 +2290,34 @@ class Search():
                 if 'corvars' in solution and v in solution['corvars']:
                     solution['corvars'] = [x for x in solution['corvars'] if x != v]
 
+    def _enforce_incompatible_specs(self, solution):
+        """Ensure at most one alt-specific variable per incompatible group.
+
+        Operates on the alt-specific dummy column names (e.g.
+        ``TIME_Car``, ``COST_Car``).  If multiple members of a group
+        are present, keep the first and remove the rest from asvars.
+        """
+        groups = self._get_incompatible_specs_groups()
+        if not groups:
+            return
+        all_v = self._all_vars_in_solution(solution)
+        for group in groups:
+            present = [v for v in group if v in all_v]
+            if len(present) <= 1:
+                continue
+            remove = set(present[1:])
+            for v in remove:
+                if 'asvars' in solution and v in solution['asvars']:
+                    solution['asvars'] = [x for x in solution['asvars'] if x != v]
+                if 'isvars' in solution and v in solution['isvars']:
+                    solution['isvars'] = [x for x in solution['isvars'] if x != v]
+                if 'randvars' in solution and v in solution['randvars']:
+                    del solution['randvars'][v]
+                if 'bcvars' in solution and v in solution['bcvars']:
+                    solution['bcvars'] = [x for x in solution['bcvars'] if x != v]
+                if 'corvars' in solution and v in solution['corvars']:
+                    solution['corvars'] = [x for x in solution['corvars'] if x != v]
+
     def _cull_attrited_vars(self):
         """Remove variables from the available pools that have exceeded the
         failure threshold.  Once removed they stay out for the rest of
@@ -2234,13 +2337,50 @@ class Search():
             print(f"  [attrition] '{v}' removed from search after {self._var_failures[v]} failures")
 
     def _apply_latent_class_constraints(self, solution, lc_constraints):
-        """Apply constraints specific to latent class models."""
-        # Example: force variables to appear only in certain classes
+        """Apply constraints specific to latent class models.
+
+        Populates solution['class_params_spec'] from the constraint builder's
+        class_specific_vars dict, so that the search respects per-class
+        variable assignments.
+        """
+        n_classes = lc_constraints.get('n_classes', 2)
+
+        # --- class-specific variable assignments ---
         if 'class_specific_vars' in lc_constraints:
-            # This would require more complex logic depending on latent class implementation
-            # For now, just ensure variables are included
-            for class_vars in lc_constraints['class_specific_vars'].values():
-                solution['asvars'] = list(set(class_vars + solution['asvars']))
+            class_spec_vars = lc_constraints['class_specific_vars']
+            # Build / update class_params_spec
+            existing = solution.data.get('class_params_spec', None)
+            if existing is None or len(existing) != n_classes:
+                existing = np.empty(n_classes, dtype=object)
+                for c in range(n_classes):
+                    existing[c] = np.array([], dtype='<U64')
+
+            for c in range(n_classes):
+                key = f'class_{c}'
+                if key in class_spec_vars:
+                    vars_c = class_spec_vars[key]
+                    current = list(existing[c]) if len(existing[c]) > 0 else []
+                    merged = list(set(current + vars_c))
+                    existing[c] = np.array(merged, dtype='<U64')
+
+            solution['class_params_spec'] = existing
+
+            # Also ensure all class-specific vars are in asvars (the flat list used
+            # by variable-selection operators for discovery)
+            all_class_vars = []
+            for vars_c in class_spec_vars.values():
+                all_class_vars.extend(vars_c)
+            solution['asvars'] = list(set(all_class_vars + solution['asvars']))
+
+        # --- membership variable constraints ---
+        if 'membership_vars' in lc_constraints:
+            member_vars = list(lc_constraints['membership_vars'].keys())
+            existing_member = solution.data.get('member_params_spec', None)
+            if existing_member is None or len(existing_member) == 0:
+                solution['member_params_spec'] = np.array(member_vars, dtype='<U64')
+            else:
+                merged = list(set(list(existing_member) + member_vars))
+                solution['member_params_spec'] = np.array(merged, dtype='<U64')
     
     def _apply_mixed_model_constraints(self, solution, mm_constraints):
         """Apply constraints specific to mixed models."""
@@ -2562,6 +2702,7 @@ class Search():
     # {
         candidate = [var for var in self.param.asvarnames if var not in solution['asvars']]
         blocked = self._get_excluded_by_mutual_group(set(solution.get('asvars', [])) | set(solution.get('isvars', [])))
+        blocked |= self._get_excluded_by_incompatible_specs(set(solution.get('asvars', [])) | set(solution.get('isvars', [])))
         candidate = [v for v in candidate if v not in blocked]
         if len(candidate) > 0:
             new_asvar = self.random_choice(candidate)
@@ -2637,6 +2778,7 @@ class Search():
     # {
         candidate = [var for var in self.param.isvarnames if var not in solution['isvars']]
         blocked = self._get_excluded_by_mutual_group(set(solution.get('asvars', [])) | set(solution.get('isvars', [])))
+        blocked |= self._get_excluded_by_incompatible_specs(set(solution.get('asvars', [])) | set(solution.get('isvars', [])))
         candidate = [v for v in candidate if v not in blocked]
         if len(candidate) > 0:
         # {
@@ -3883,47 +4025,72 @@ class Search():
     def create_dummy_column(self, asvars):
     # {
         """
-        This function generates a random boolean array with the same
-        length as 'asvars'. For each 'True' in the array, it creates a
-        new dataframe column for each choice alternative where the value
-        is the product of the variable and a boolean that indicates if
-        the choice alternative is the current alternative. It then adds
-        these alternative-specific variables into a new list and extends
-        it with the remaining variables from 'asvars' that were not
-        chosen for alternative-specific coefficients.
-
-        The new asvar includes the new dummy columns created and the remaining
-        variables from 'asvars' that were not chosen for alternative-specific coefficients.
+        Generates alternative-specific dummy columns for variables that
+        receive alternative-specific coefficients.  For variables listed
+        in the ``asvar_alt_spec`` constraint, dummies are created *only*
+        for the specified alternatives (the original variable is kept
+        for the remaining alternatives as a generic coefficient).  For
+        all other variables the existing random-selection behaviour is
+        preserved: a random subset is turned into full alt-specific
+        dummies and the remaining variables stay generic.
+
+        The new asvar list includes the new dummy columns alongside the
+        variables that were not chosen for alternative-specific
+        coefficients.
         """
 
-        # Generate a random boolean array with the same length as asvars
-        rand_array = np.random.choice([True, False], len(asvars))
+        constraints = getattr(self.param, 'pres_spec_constr', None) or {}
+        alt_spec_map = constraints.get('asvar_alt_spec', {})
+
+        # Separate constrained vars from free vars
+        constrained_asvars = [v for v in asvars if v in alt_spec_map]
+        free_asvars = [v for v in asvars if v not in alt_spec_map]
+
+        # Generate a random boolean array for free vars only
+        rand_array = np.random.choice([True, False], len(free_asvars))
 
-        # Initialize a list for storing new alternative-specific variables
         asvars_new = []
 
-        # Extract variables randomly chosen to have alternative-specific coefficients
-        alt_spec_vars = [var for var, bool in zip(asvars, rand_array) if bool]
+        # 1. Process constrained variables (subset alt-specific)
+        for alt_var in constrained_asvars:
+        # {
+            specified_alts = alt_spec_map[alt_var]
+            for choice in specified_alts:
+            # {
+                col_name = f"{alt_var}_{choice}"
+                if col_name not in self.param.df.columns:
+                    self.param.df[col_name] = self.param.df[alt_var] * (self.param.alt_var == choice)
+
+                    if self.param.nb_crit > 1:
+                        self.param.df_test[col_name] = self.param.df_test[alt_var] * (self.param.test_alt_var == choice)
+
+                asvars_new.append(col_name)
+            # }
+            # Keep the original variable as a generic coefficient for
+            # the alternatives NOT listed in the constraint
+            asvars_new.append(alt_var)
+        # }
+
+        # 2. Process free variables (existing random behaviour)
+        alt_spec_vars = [var for var, bool in zip(free_asvars, rand_array) if bool]
+        generic_vars  = [var for var, bool in zip(free_asvars, rand_array) if not bool]
 
-        # Create dummy columns for the selected variables
         for alt_var in alt_spec_vars:
         # {
             for choice in self.param.choice_set:
             # {
                 col_name = f"{alt_var}_{choice}"
-                self.param.df[col_name] = self.param.df[alt_var] * (self.param.alt_var == choice)
+                if col_name not in self.param.df.columns:
+                    self.param.df[col_name] = self.param.df[alt_var] * (self.param.alt_var == choice)
 
-                if self.param.nb_crit > 1:
-                    # Must compare against the split TEST alt array, not the
-                    # full training self.param.alt_var (shape mismatch else).
-                    self.param.df_test[col_name] = self.param.df_test[alt_var] * (self.param.test_alt_var == choice)
+                    if self.param.nb_crit > 1:
+                        self.param.df_test[col_name] = self.param.df_test[alt_var] * (self.param.test_alt_var == choice)
 
                 asvars_new.append(col_name)
             # }
         # }
+        asvars_new.extend(generic_vars)
 
-        # Add the remaining variables that were not selected for alternative-specific coefficients
-        asvars_new.extend(var for var, bool in zip(asvars, rand_array) if not bool)
         return asvars_new
 
     # }
@@ -4148,6 +4315,7 @@ class Search():
             avail=avail,
             membership_vars=membership_vars,
             member_params_spec=member_params_spec,
+            class_params_spec=class_params_spec,
         )
         model.fit(em_method="squarem")
         model.get_loglik_null()
@@ -4344,6 +4512,29 @@ class Search():
         member_params_spec = sol.get('member_params_spec', None)
         num_classes = getattr(self.param, 'num_classes', 2)
 
+        # Build all_vars as the UNION of all per-class variable specs
+        # (plus membership vars, plus asvars/isvars fallback).
+        if class_params_spec is not None and len(class_params_spec) > 0:
+            all_vars_set = set()
+            for c in range(len(class_params_spec)):
+                if class_params_spec[c] is not None and len(class_params_spec[c]) > 0:
+                    for v in class_params_spec[c]:
+                        all_vars_set.add(str(v))
+            # Also include member params
+            if member_params_spec is not None:
+                if hasattr(member_params_spec, 'flat'):
+                    for v in member_params_spec.flat:
+                        all_vars_set.add(str(v))
+                elif hasattr(member_params_spec, '__iter__'):
+                    for v in member_params_spec:
+                        all_vars_set.add(str(v))
+            # Fallback to asvars+isvars for any vars not captured
+            for v in (as_vars + is_vars):
+                all_vars_set.add(str(v))
+            all_vars = [v for v in self.param.varnames if v in all_vars_set]
+        else:
+            all_vars = [var for var in self.param.varnames if var in (as_vars + is_vars)]
+
         X = self.param.df[all_vars].values
         y = self.param.choices
         ids = self.param.choice_id if self.param.choice_id is not None else self.param.ind_id

searchlibrium-0.0.152/src/SearchLibrium/version.txt

@@ -0,0 +1 @@
+0.0.152

{searchlibrium-0.0.151 → searchlibrium-0.0.152/src/SearchLibrium.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: SearchLibrium
-Version: 0.0.151
+Version: 0.0.152
 Summary: A Python package for econometric models driven by search
 Author: Mario Fernando Taco Morales, Alexander Paz, Prithvi Beeramole, Robert Burdett
 Author-email: Zeke Ahern <z.ahern@qut.edu.au>

{searchlibrium-0.0.151 → searchlibrium-0.0.152}/src/SearchLibrium.egg-info/SOURCES.txt

@@ -53,4 +53,5 @@ src/SearchLibrium.egg-info/top_level.txt
 src/SearchLibrium/bhhh/minimize.py
 src/SearchLibrium/data/Swissmetro_final.csv
 src/SearchLibrium/data/TravelMode.csv
-src/SearchLibrium/data/electricity.csv
+src/SearchLibrium/data/electricity.csv
+src/data_generator/latent_class_gen.py

{searchlibrium-0.0.151 → searchlibrium-0.0.152}/src/SearchLibrium.egg-info/top_level.txt

@@ -1,2 +1,3 @@
 SearchLibrium
+data_generator
 sa_runs

searchlibrium-0.0.152/src/data_generator/latent_class_gen.py

@@ -0,0 +1,174 @@
+import numpy as np
+import pandas as pd
+from scipy.special import softmax
+
+class AdvancedLatentClassGenerator:
+    def __init__(
+        self,
+        n_classes=3,
+        n_alternatives=3,
+        n_individuals=2000,
+        n_choice_tasks=5,
+        class_probs=None,
+        scale_separation=2.5,
+        n_noise_vars=5,
+        n_weak_vars=3,
+        n_collinear_vars=3,
+        rare_class=False,
+        random_state=42
+    ):
+        self.rng = np.random.default_rng(random_state)
+
+        self.K = n_classes
+        self.J = n_alternatives
+        self.N = n_individuals
+        self.T = n_choice_tasks
+
+        if class_probs is None:
+            if rare_class:
+                probs = np.ones(n_classes)
+                probs[-1] = 0.05
+                probs[:-1] = (1 - 0.05) / (n_classes - 1)
+                self.class_probs = probs
+            else:
+                self.class_probs = np.ones(n_classes) / n_classes
+        else:
+            self.class_probs = np.array(class_probs)
+
+        self.scale_separation = scale_separation
+        self.n_noise_vars = n_noise_vars
+        self.n_weak_vars = n_weak_vars
+        self.n_collinear_vars = n_collinear_vars
+
+        self._define_variables()
+        self._generate_parameters()
+
+    # --------------------------------
+    # Variable Names
+    # --------------------------------
+    def _define_variables(self):
+
+        self.shared_vars = [
+            "price",
+            "travel_time",
+            "waiting_time"
+        ]
+
+        self.class_specific_vars = {
+            0: ["comfort_level", "seat_space"],
+            1: ["brand_reputation", "loyalty_points"],
+            2: ["eco_rating", "carbon_emissions"]
+        }
+
+        self.noise_vars = [f"random_noise_{i}" for i in range(self.n_noise_vars)]
+        self.weak_vars = [f"weak_signal_{i}" for i in range(self.n_weak_vars)]
+        self.collinear_vars = [f"collinear_var_{i}" for i in range(self.n_collinear_vars)]
+
+    # --------------------------------
+    # Generate True Parameters
+    # --------------------------------
+    def _generate_parameters(self):
+
+        self.parameters = {}
+
+        for k in range(self.K):
+            params = {}
+
+            # Shared variables
+            for var in self.shared_vars:
+                params[var] = self.rng.normal(
+                    loc=k * self.scale_separation,
+                    scale=1.0
+                )
+
+            # Class-specific variables
+            for var in self.class_specific_vars.get(k, []):
+                params[var] = self.rng.normal(
+                    loc=(k + 1) * self.scale_separation,
+                    scale=1.0
+                )
+
+            # Weak effect variables
+            for var in self.weak_vars:
+                params[var] = self.rng.normal(
+                    loc=0,
+                    scale=0.05
+                )
+
+            # Noise variables → TRUE beta = 0
+            for var in self.noise_vars:
+                params[var] = 0.0
+
+            self.parameters[k] = params
+
+    # --------------------------------
+    # Feature Generation
+    # --------------------------------
+    def _generate_features(self):
+
+        all_vars = set(self.shared_vars)
+        for v in self.class_specific_vars.values():
+            all_vars.update(v)
+
+        all_vars.update(self.noise_vars)
+        all_vars.update(self.weak_vars)
+        all_vars.update(self.collinear_vars)
+
+        X = {}
+
+        for var in all_vars:
+            X[var] = self.rng.normal(size=(self.N, self.T, self.J))
+
+        # Create collinearity
+        for i, var in enumerate(self.collinear_vars):
+            X[var] = 0.8 * X["price"] + 0.2 * self.rng.normal(
+                size=(self.N, self.T, self.J)
+            )
+
+        return X
+
+    # --------------------------------
+    # Data Generation
+    # --------------------------------
+    def generate(self):
+
+        classes = self.rng.choice(
+            self.K,
+            size=self.N,
+            p=self.class_probs
+        )
+
+        X = self._generate_features()
+        rows = []
+
+        for n in range(self.N):
+            k = classes[n]
+
+            for t in range(self.T):
+
+                utilities = np.zeros(self.J)
+
+                for j in range(self.J):
+                    for var, beta in self.parameters[k].items():
+                        utilities[j] += beta * X[var][n, t, j]
+
+                probs = softmax(utilities)
+                choice = self.rng.choice(self.J, p=probs)
+
+                for j in range(self.J):
+
+                    row = {
+                        "individual": n,
+                        "choice_task": t,
+                        "alternative": j,
+                        "choice": 1 if j == choice else 0,
+                        "true_class": k
+                    }
+
+                    for var in X:
+                        row[var] = X[var][n, t, j]
+
+                    rows.append(row)
+
+        df = pd.DataFrame(rows)
+        return df

searchlibrium-0.0.151/src/SearchLibrium/version.txt

@@ -1 +0,0 @@
-0.0.151