diff --git a/osl_dynamics/analysis/spectral.py b/osl_dynamics/analysis/spectral.py
index bc1bd96c..40094a06 100644
--- a/osl_dynamics/analysis/spectral.py
+++ b/osl_dynamics/analysis/spectral.py
@@ -1406,10 +1406,10 @@ def _welch(
     # Rescale PSDs to account for the number of time points
     # each state was active
     fo = np.sum(stc, axis=0) / stc.shape[0]
-    for i, (psd_, fo_) in enumerate(zip(psd, fo)):
-        psd_ /= fo_
-        if np.isnan(psd_).any():
-            psd[i] = np.nan_to_num(psd_)  # zero out nan values
+    for i in range(len(fo)):
+        psd[i] /= fo[i]
+        if np.isnan(psd[i]).any():
+            psd[i] = np.nan_to_num(psd[i])  # zero out nan values
             _logger.warn(
                 "PSD contains NaN values. This may indicate a potentially "
                 "poor HMM fit. You should consider running the model again "
@@ -1714,7 +1714,6 @@ def _multitaper(
 
         if calc_coh:
             # Create a channels by channels matrix for cross PSDs
-            n_channels = data.shape[-1]
             n_freq = p.shape[-1]
             cpsd = np.empty(
                 [n_channels, n_channels, n_freq],
@@ -1734,10 +1733,10 @@ def _multitaper(
     # Rescale PSDs to account for the number of time points
     # each state was active
     fo = np.sum(stc, axis=0) / stc.shape[0]
-    for i, (psd_, fo_) in enumerate(zip(psd, fo)):
-        psd_ /= fo_
-        if np.isnan(psd_).any():
-            psd[i] = np.nan_to_num(psd_)  # zero out nan values
+    for i in range(len(fo)):
+        psd[i] /= fo[i]
+        if np.isnan(psd[i]).any():
+            psd[i] = np.nan_to_num(psd[i])  # zero out nan values
             _logger.warn(
                 "PSD contains NaN values. This may indicate a potentially "
                 "poor HMM fit. You should consider running the model again "