Update to return p-values

voytekresearch · Jan 11, 2020 · cf69446 · cf69446
1 parent 54355e7
commit cf69446
Show file tree

Hide file tree

Showing 3 changed files with 49 additions and 40 deletions.
diff --git a/bratios/analysis.py b/bratios/analysis.py
@@ -86,15 +86,17 @@ def get_wave_params(band_label):
     return [curr_band + "_" + feat for feat in FEATURE_LABELS]
 
 
-def print_aperiodic_correlation(ratio_type, corr, show=True):
+def print_aperiodic_correlation(ratio_type, corr, ps=None, show=True):
     """Prints out the correlation between ratio and aperiodic parameters.
 
     Parameters
     ----------
     ratio_type : string
         Which specific ratio measure to report (ex. "TBR").
-    corr : list of floats
+    corr : list of float
         Correlation r-values for aperiodic params.
+    ps : list of float
+        P-values for the correlations.
     show : boolean
         Whether to display the print out.
 
@@ -105,7 +107,8 @@ def print_aperiodic_correlation(ratio_type, corr, show=True):
 
     if show:
         for ind, param in enumerate(["Exp", "Off", "Age"]):
-            print("The corr of {} to {} is {:1.2f}".format(ratio_type, param, corr[ind]))
+            p_txt = "\t with a p-value of {:1.4f}".format(ps[ind]) if isinstance(ps, np.ndarray) else ""
+            print(("The corr of {} to {} is {:1.2f}" + p_txt).format(ratio_type, param, corr[ind]))
 
 
 def get_non_ratio_band(ratio_type):
@@ -139,8 +142,7 @@ def param_ratio_corr(df, ratio_type, ch_inds, func=nan_corr_pearson):
     df : 2D DataFrame
         Container which holds ratio, channels, and peak values.
     ratio_type : string
-        Ratio measure to run correlations across.
-        ex) "TBR"
+        Ratio measure to run correlations across. ex: "TBR".
     ch_inds : list of ints
         Channels to run correlations over.
     func : callable
@@ -163,23 +165,27 @@ def param_ratio_corr(df, ratio_type, ch_inds, func=nan_corr_pearson):
     alpha = get_wave_params("A")
     beta = get_wave_params("B")
 
-    theta_corrs = np.zeros(3)
-    alpha_corrs = np.zeros(3)
-    beta_corrs = np.zeros(3)
+    # Initialize variables to store periodic & aperiodic correlation results
+    per_corrs = np.zeros([3, 3])
+    per_ps = np.zeros([3, 3])
     ap_corrs = np.zeros(3)
+    ap_ps = np.zeros(3)
 
     # Ratio vs spectral params correlation
     for ind in range(3):
-        theta_corrs[ind] = func(rel_df[theta[ind]], rel_df[ratio_type])[0]
-        alpha_corrs[ind] = func(rel_df[alpha[ind]], rel_df[ratio_type])[0]
-        beta_corrs[ind] = func(rel_df[beta[ind]], rel_df[ratio_type])[0]
+        per_corrs[0, ind], per_ps[0, ind] = \
+            func(rel_df[theta[ind]], rel_df[ratio_type])
+        per_corrs[1, ind], per_ps[1, ind] = \
+            func(rel_df[alpha[ind]], rel_df[ratio_type])
+        per_corrs[2, ind], per_ps[2, ind] = \
+            func(rel_df[beta[ind]], rel_df[ratio_type])
 
     # Ratio vs aperiodic params correlation
-    ap_corrs[0] = func(rel_df["Exp"], rel_df[ratio_type])[0]
-    ap_corrs[1] = func(rel_df["Off"], rel_df[ratio_type])[0]
-    ap_corrs[2] = func(rel_df["Age"], rel_df[ratio_type])[0]
+    ap_corrs[0], ap_ps[0] = func(rel_df["Exp"], rel_df[ratio_type])
+    ap_corrs[1], ap_ps[1] = func(rel_df["Off"], rel_df[ratio_type])
+    ap_corrs[2], ap_ps[2] = func(rel_df["Age"], rel_df[ratio_type])
 
-    return np.stack((theta_corrs, alpha_corrs,beta_corrs)), ap_corrs
+    return per_corrs, ap_corrs, per_ps, ap_ps
 
 
 def _add_params(curr_row, theta_params, beta_params, alpha_params, ap):
@@ -361,8 +367,7 @@ def param_corr(df, corr_label_1, corr_label_2, chan_inds, func):
 
     Returns
     -------
-
-    float describing results from correlation function.
+    R-value and p-value from the correlation function.
     """
 
     # Select relevant rows from df
@@ -371,4 +376,4 @@ def param_corr(df, corr_label_1, corr_label_2, chan_inds, func):
     # Average across selected channels per subject
     rel_df = rel_df.groupby("Subj_ID").mean()
 
-    return func(rel_df[corr_label_1], rel_df[corr_label_2])[0]
+    return func(rel_df[corr_label_1], rel_df[corr_label_2])
diff --git a/notebooks/5-EEG-Analysis.ipynb b/notebooks/5-EEG-Analysis.ipynb
@@ -427,7 +427,8 @@
    "outputs": [],
    "source": [
     "# Calculate the correlation between ratios and spectral features\n",
-    "tbr_per, tbr_ap = param_ratio_corr(df, \"TBR\", all_chans, func=nan_corr_spearman)"
+    "tbr_per, tbr_ap, tbr_per_ps, tbr_ap_ps = \\\n",
+    "    param_ratio_corr(df, \"TBR\", all_chans, func=nan_corr_spearman)"
    ]
   },
   {
@@ -494,7 +495,7 @@
    "outputs": [],
    "source": [
     "# Frontal Channels\n",
-    "tbr_per_fr, tbr_ap_fr = param_ratio_corr(df, \"TBR\", ch_inds_frontal, func=nan_corr_spearman)\n",
+    "tbr_per_fr, tbr_ap_fr, _, _ = param_ratio_corr(df, \"TBR\", ch_inds_frontal, func=nan_corr_spearman)\n",
     "print_aperiodic_correlation(\"TBR\", tbr_ap_fr, show=SHOW_SUB)\n",
     "\n",
     "plot_param_ratio_corr(tbr_per_fr, show=SHOW_SUB,\n",
@@ -510,7 +511,7 @@
    "outputs": [],
    "source": [
     "# Central channels\n",
-    "tbr_per_ce, tbr_ap_ce = param_ratio_corr(df, \"TBR\", ch_inds_central, func=nan_corr_spearman)\n",
+    "tbr_per_ce, tbr_ap_ce, _, _ = param_ratio_corr(df, \"TBR\", ch_inds_central, func=nan_corr_spearman)\n",
     "print_aperiodic_correlation(\"TBR\", tbr_ap_ce, show=SHOW_SUB)\n",
     "\n",
     "plot_param_ratio_corr(tbr_per_ce, show=SHOW_SUB,\n",
@@ -526,7 +527,7 @@
    "outputs": [],
    "source": [
     "# Parietal channels\n",
-    "tbr_per_pa, tbr_ap_pa = param_ratio_corr(df, \"TBR\", ch_inds_parietal, func=nan_corr_spearman)\n",
+    "tbr_per_pa, tbr_ap_pa, _, _ = param_ratio_corr(df, \"TBR\", ch_inds_parietal, func=nan_corr_spearman)\n",
     "print_aperiodic_correlation(\"TBR\", tbr_ap_pa, show=SHOW_SUB)\n",
     "\n",
     "plot_param_ratio_corr(tbr_per_pa, show=SHOW_SUB,\n",
@@ -560,7 +561,8 @@
    "outputs": [],
    "source": [
     "# Calculate the correlation between ratios and spectral features\n",
-    "tar_per, tar_ap = param_ratio_corr(df, \"TAR\", all_chans, func=nan_corr_spearman)"
+    "tar_per, tar_ap, tar_per_ps, tar_ap_ps = \\\n",
+    "    param_ratio_corr(df, \"TAR\", all_chans, func=nan_corr_spearman)"
    ]
   },
   {
@@ -627,7 +629,7 @@
    "outputs": [],
    "source": [
     "# Frontal Channels\n",
-    "tar_per_fr, tar_ap_fr = param_ratio_corr(df, \"TAR\", ch_inds_frontal, func=nan_corr_spearman)\n",
+    "tar_per_fr, tar_ap_fr, _, _ = param_ratio_corr(df, \"TAR\", ch_inds_frontal, func=nan_corr_spearman)\n",
     "print_aperiodic_correlation(\"TAR\", tar_ap_fr, show=SHOW_SUB)\n",
     "\n",
     "plot_param_ratio_corr(tar_per_fr, show=SHOW_SUB,\n",
@@ -643,7 +645,7 @@
    "outputs": [],
    "source": [
     "# Central Channels\n",
-    "tar_per_ce, tar_ap_ce = param_ratio_corr(df, \"TAR\", ch_inds_central, func=nan_corr_spearman)\n",
+    "tar_per_ce, tar_ap_ce, _, _ = param_ratio_corr(df, \"TAR\", ch_inds_central, func=nan_corr_spearman)\n",
     "print_aperiodic_correlation(\"TAR\", tar_ap_ce, show=SHOW_SUB)\n",
     "\n",
     "plot_param_ratio_corr(tar_per_ce, show=SHOW_SUB,\n",
@@ -659,7 +661,7 @@
    "outputs": [],
    "source": [
     "# Parietal Channels\n",
-    "tar_per_pa, tar_ap_pa = param_ratio_corr(df, \"TAR\", ch_inds_parietal, func=nan_corr_spearman)\n",
+    "tar_per_pa, tar_ap_pa, _, _ = param_ratio_corr(df, \"TAR\", ch_inds_parietal, func=nan_corr_spearman)\n",
     "print_aperiodic_correlation(\"TAR\", tar_ap_pa, show=SHOW_SUB)\n",
     "\n",
     "plot_param_ratio_corr(tar_per_pa, show=SHOW_SUB,\n",
@@ -693,7 +695,8 @@
    "outputs": [],
    "source": [
     "# Calculate the correlation between ratios and spectral features\n",
-    "abr_per, abr_ap = param_ratio_corr(df, \"ABR\", all_chans, func=nan_corr_spearman)"
+    "abr_per, abr_ap, abr_per_ps, abr_ap_ps = \\\n",
+    "    param_ratio_corr(df, \"ABR\", all_chans, func=nan_corr_spearman)"
    ]
   },
   {
@@ -760,7 +763,7 @@
    "outputs": [],
    "source": [
     "# Frontal Channels\n",
-    "abr_per_fr, abr_ap_fr = param_ratio_corr(df, \"ABR\", ch_inds_frontal, func=nan_corr_spearman)\n",
+    "abr_per_fr, abr_ap_fr, _, _ = param_ratio_corr(df, \"ABR\", ch_inds_frontal, func=nan_corr_spearman)\n",
     "print_aperiodic_correlation(\"ABR\", abr_ap_fr, show=SHOW_SUB)\n",
     "\n",
     "plot_param_ratio_corr(abr_per_fr, show=SHOW_SUB,\n",
@@ -776,7 +779,7 @@
    "outputs": [],
    "source": [
     "# Central Channels\n",
-    "abr_per_ce, abr_ap_ce = param_ratio_corr(df, \"ABR\", ch_inds_central, func=nan_corr_spearman)\n",
+    "abr_per_ce, abr_ap_ce, _, _ = param_ratio_corr(df, \"ABR\", ch_inds_central, func=nan_corr_spearman)\n",
     "print_aperiodic_correlation(\"ABR\", abr_ap_ce, show=SHOW_SUB)\n",
     "\n",
     "plot_param_ratio_corr(abr_per_ce, show=SHOW_SUB,\n",
@@ -792,7 +795,7 @@
    "outputs": [],
    "source": [
     "# Parietal Channels\n",
-    "abr_per_pa, abr_ap_pa = param_ratio_corr(df, \"ABR\", ch_inds_parietal, func=nan_corr_spearman)\n",
+    "abr_per_pa, abr_ap_pa, _, _ = param_ratio_corr(df, \"ABR\", ch_inds_parietal, func=nan_corr_spearman)\n",
     "print_aperiodic_correlation(\"ABR\", abr_ap_pa, show=SHOW_SUB)\n",
     "\n",
     "plot_param_ratio_corr(abr_per_pa, show=SHOW_SUB,\n",
@@ -825,8 +828,8 @@
    ],
    "source": [
     "# Check correlation across all channels\n",
-    "print('Age-EXP corr  global: \\t\\t {:1.4f}'.format(\n",
-    "    param_corr(df, \"Age\", \"Exp\", all_chans, nan_corr_spearman)))"
+    "age_exp_corr, age_exp_p = param_corr(df, \"Age\", \"Exp\", all_chans, nan_corr_spearman)\n",
+    "print('Age-EXP corr  global: \\t\\t {:1.4f}'.format(age_exp_corr))"
    ]
   },
   {
@@ -847,11 +850,11 @@
    "source": [
     "# Check correlation of age and exponent within channel sub-selections\n",
     "print('Age-EXP corr  frontal: \\t\\t {:1.4f}'.format(\n",
-    "      param_corr(df, \"Age\", \"Exp\", ch_inds_frontal, nan_corr_spearman)))\n",
+    "      param_corr(df, \"Age\", \"Exp\", ch_inds_frontal, nan_corr_spearman)[0]))\n",
     "print('Age-EXP corr  central: \\t\\t {:1.4f}'.format(\n",
-    "      param_corr(df, \"Age\", \"Exp\", ch_inds_central, nan_corr_spearman)))\n",
+    "      param_corr(df, \"Age\", \"Exp\", ch_inds_central, nan_corr_spearman)[0]))\n",
     "print('Age-EXP corr  parietal: \\t {:1.4f}'.format(\n",
-    "      param_corr(df, \"Age\", \"Exp\", ch_inds_parietal, nan_corr_spearman)))"
+    "      param_corr(df, \"Age\", \"Exp\", ch_inds_parietal, nan_corr_spearman)[0]))"
    ]
   }
  ],

diff --git a/notebooks/6-EEG-Topographies.ipynb b/notebooks/6-EEG-Topographies.ipynb
@@ -69,7 +69,7 @@
    "outputs": [],
    "source": [
     "# Settings\n",
-    "SAVE_FIG = True"
+    "SAVE_FIG = False"
    ]
   },
   {
@@ -568,11 +568,12 @@
    "outputs": [],
    "source": [
     "corr_mat = np.zeros((3, 4))\n",
+    "ps_mat = np.zeros((3, 4))\n",
     "for ind, ratio in enumerate([tbrs, tars, abrs]):\n",
-    "    corr_mat[ind][0] = spearmanr(ratio, theta_pw)[0]\n",
-    "    corr_mat[ind][1] = spearmanr(ratio, alpha_pw)[0]\n",
-    "    corr_mat[ind][2] = spearmanr(ratio, beta_pw)[0]\n",
-    "    corr_mat[ind][3] = spearmanr(ratio, exps)[0]"
+    "    corr_mat[ind][0], ps_mat[ind][0] = spearmanr(ratio, theta_pw)\n",
+    "    corr_mat[ind][1], ps_mat[ind][1] = spearmanr(ratio, alpha_pw)\n",
+    "    corr_mat[ind][2], ps_mat[ind][2] = spearmanr(ratio, beta_pw)\n",
+    "    corr_mat[ind][3], ps_mat[ind][3] = spearmanr(ratio, exps)"
    ]
   },
   {