verify negative binomial objective function

src/pySODM/optimization/objective_functions.py

@@ -86,21 +86,13 @@ def ll_negative_binomial(ymodel, ydata, alpha):
     ll: float
         Loglikelihood belonging to the comparison of the data points and the model prediction.
     """
-
+    
     # Expand first dimensions on 'alpha' to match the axes
     alpha = np.array(alpha)[np.newaxis, ...]
-    # Convert datatype to float
-    ymodel = ymodel.astype('float64')
-    ydata = ydata.astype('float64')
-    # Raise ymodel or ydata if there are negative values present
-    if ((np.min(ymodel) < 0) | (np.min(ydata) < 0)):
-        offset_value = (-1 - 1e-6)*np.min([np.min(ymodel), np.min(ydata)])
-        ymodel += offset_value
-        ydata += offset_value
-        #warnings.warn(f"One or more values in the prediction were negative thus the prediction was offset, minimum predicted value: {min(ymodel)}")
-    elif ((np.min(ymodel) == 0) | (np.min(ydata) == 0)):
-        ymodel += 1e-6
-        ydata += 1e-6
+
+    # Convert datatype to float and add one
+    ymodel = ymodel.astype('float64') + 1 
+    ydata = ydata.astype('float64') + 1
 
     return np.sum(ydata*np.log(ymodel)) - np.sum((ydata + 1/alpha)*np.log(1+alpha*ymodel)) + np.sum(ydata*np.log(alpha)) + np.sum(gammaln(ydata+1/alpha)) - np.sum(gammaln(ydata+1))- np.sum(ydata.shape[0]*gammaln(1/alpha))