pyfixest 0.9.11 (#165)

docs/news.md

@@ -1,5 +1,10 @@
 # News
 
+## PyFixest `0.9.11`
+
+- Bump required `formulaic` version to `0.6.5`.
+- Stop copying the data frame in `fixef()`.
+
 ## PyFixest `0.9.10`
 
 - Fixes a big in the `wildboottest` method (see [#158](https://github.com/s3alfisc/pyfixest/issues/158)).

pyfixest/feols.py

@@ -719,7 +719,7 @@ def fixef(self) -> None:
         This method creates the following attributes:
 
         - `alphaDF` (pd.DataFrame): A DataFrame with the estimated fixed effects.
-        - `sumDF` (np.array): An array with the sum of fixed effects for each observation (i = 1, ..., N).
+        - `sumFE` (np.array): An array with the sum of fixed effects for each observation (i = 1, ..., N).
 
         Args:
             None
@@ -747,19 +747,18 @@ def fixef(self) -> None:
         depvars, res = _fml.split("~")
         covars, fixef_vars = res.split("|")
 
-        df = _data.copy()
-        # all fixef vars to pd.Categorical
-        for x in fixef_vars.split("+"):
-            df[x] = pd.Categorical(df[x])
+        fixef_vars = fixef_vars.split("+")
+        fixef_vars_C = [f"C({x})" for x in fixef_vars]
+        fixef_fml = "+".join(fixef_vars_C)
 
         fml_linear = f"{depvars} ~ {covars}"
-        Y, X = model_matrix(fml_linear, df)
+        Y, X = model_matrix(fml_linear, _data)
         X = X[self._coefnames]  # drop intercept, potentially multicollinear vars
         Y = Y.to_numpy().flatten().astype(np.float64)
         X = X.to_numpy()
         uhat = csr_matrix(Y - X @ self._beta_hat).transpose()
 
-        D2 = model_matrix("-1+" + fixef_vars, df, output="sparse")
+        D2 = model_matrix("-1+" + fixef_fml, _data, output="sparse")
         cols = D2.model_spec.column_names
 
         alpha = spsolve(D2.transpose() @ D2, D2.transpose() @ uhat)
@@ -829,7 +828,6 @@ def predict(self, newdata: Optional[pd.DataFrame] = None) -> np.ndarray:
                     self.fixef()
 
                 fvals = self._fixef.split("+")
-
                 df_fe = newdata[fvals].astype(str)
 
                 # populate matrix with fixed effects estimates
@@ -839,7 +837,9 @@ def predict(self, newdata: Optional[pd.DataFrame] = None) -> np.ndarray:
                 for i, fixef in enumerate(df_fe.columns):
                     new_levels = df_fe[fixef].unique()
                     old_levels = _data[fixef].unique().astype(str)
-                    subdict = self._fixef_dict[fixef]
+                    subdict = self._fixef_dict[
+                        f"C({fixef})"
+                    ]  # as variables are called C(var) in the fixef_dict
 
                     for level in new_levels:
                         # if level estimated: either estimated value (or 0 for reference level)

pyproject.toml

@@ -1,9 +1,9 @@
 [tool.poetry]
 name = "pyfixest"
-version = "0.9.10"
+version = "0.9.11"
 
 
-description = "Experimental draft package for high dimensional fixed effect estimation. Supports OLS and IV estimation."
+description = "(Not so) experimental draft package for high dimensional fixed effect estimation. Supports OLS, IV and Poisson regression and a range of inference procedures."
 authors = ["Alexander Fischer <alexander-fischer1801@t-online.de>"]
 license = "MIT"
 readme = "readme.md"
@@ -16,7 +16,7 @@ pandas=">=1.1.0"
 numpy=">=1.19.0"
 PyHDFE=">=0.2"
 scipy=">=1.6"
-formulaic=">=0.6.0"
+formulaic=">=0.6.5"
 pytest="^7.0.0"
 lets-plot = "^4.0.1"
 wildboottest={ version = ">=0.2", optional = true}