Sync changes to nbdev notebooks

Author: Jacobluke-

nbs/API/confint_2group_diff.ipynb

@@ -213,24 +213,26 @@
     "\n",
     "    return out\n",
     "\n",
-    "@njit(cache=True) # parallelization must be turned off for random number generation\n",
-    "def delta2_bootstrap_loop(x1, x2, x3, x4, resamples, pooled_sd, rng_seed, is_paired):\n",
+    "\n",
+    "@njit(cache=True)\n",
+    "def delta2_bootstrap_loop(x1, x2, x3, x4, resamples, pooled_sd, rng_seed, is_paired, proportional=False):\n",
+    "    \"\"\"\n",
+    "    Compute bootstrapped differences for delta-delta, handling both regular and proportional data\n",
+    "    \"\"\"\n",
     "    np.random.seed(rng_seed)\n",
-    "    out_delta_g = np.empty(resamples)\n",
     "    deltadelta = np.empty(resamples)\n",
+    "    out_delta_g = np.empty(resamples)\n",
     "    \n",
     "    n1, n2, n3, n4 = len(x1), len(x2), len(x3), len(x4)\n",
-    "    if is_paired:\n",
-    "        if n1 != n2 or n3 != n4:\n",
-    "            raise ValueError(\"Each control group must have the same length as its corresponding test group in paired analysis.\")\n",
-    "    \n",
+    "    if is_paired and (n1 != n2 or n3 != n4):\n",
+    "        raise ValueError(\"Each control group must have the same length as its corresponding test group in paired analysis.\")\n",
     "\n",
     "    # Bootstrapping\n",
     "    for i in range(resamples):\n",
     "        # Paired or unpaired resampling\n",
     "        if is_paired:\n",
-    "            indices_1 = np.random.choice(len(x1),len(x1))\n",
-    "            indices_2 = np.random.choice(len(x3),len(x3))\n",
+    "            indices_1 = np.random.choice(len(x1), len(x1))\n",
+    "            indices_2 = np.random.choice(len(x3), len(x3))\n",
     "            x1_sample, x2_sample = x1[indices_1], x2[indices_1]\n",
     "            x3_sample, x4_sample = x3[indices_2], x4[indices_2]\n",
     "        else:\n",
@@ -241,13 +243,14 @@
     "            x1_sample, x2_sample = x1[indices_1], x2[indices_2]\n",
     "            x3_sample, x4_sample = x3[indices_3], x4[indices_4]\n",
     "\n",
-    "        # Calculating deltas\n",
+    "        # Calculate deltas\n",
     "        delta_1 = np.mean(x2_sample) - np.mean(x1_sample)\n",
     "        delta_2 = np.mean(x4_sample) - np.mean(x3_sample)\n",
     "        delta_delta = delta_2 - delta_1\n",
-    "\n",
+    "        \n",
     "        deltadelta[i] = delta_delta\n",
-    "        out_delta_g[i] = delta_delta / pooled_sd\n",
+    "\n",
+    "        out_delta_g[i] = delta_delta if proportional else delta_delta/pooled_sd\n",
     "\n",
     "    return out_delta_g, deltadelta\n",
     "\n",
@@ -258,39 +261,42 @@
     "    x3: np.ndarray,  # Control group 2\n",
     "    x4: np.ndarray,  # Test group 2\n",
     "    is_paired: str = None,\n",
-    "    resamples: int = 5000,  # The number of bootstrap resamples to be taken for the calculation of the confidence interval limits.\n",
-    "    random_seed: int = 12345,  # `random_seed` is used to seed the random number generator during bootstrap resampling. This ensures that the confidence intervals reported are replicable.\n",
-    ") -> (\n",
-    "    tuple\n",
-    "):  # bootstraped result and empirical result of deltas' g, and the bootstraped result of delta-delta\n",
+    "    resamples: int = 5000,\n",
+    "    random_seed: int = 12345,\n",
+    "    proportional: bool = False\n",
+    ") -> tuple:\n",
     "    \"\"\"\n",
-    "    Bootstraps the effect size deltas' g.\n",
-    "\n",
+    "    Bootstraps the effect size deltas' g or proportional delta-delta\n",
     "    \"\"\"\n",
-    "\n",
     "    x1, x2, x3, x4 = map(np.asarray, [x1, x2, x3, x4])\n",
-    "\n",
-    "    # Calculating pooled sample standard deviation\n",
-    "    stds = [np.std(x) for x in [x1, x2, x3, x4]]\n",
-    "    ns = [len(x) for x in [x1, x2, x3, x4]]\n",
-    "\n",
-    "    sd_numerator = sum((n - 1) * s**2 for n, s in zip(ns, stds))\n",
-    "    sd_denominator = sum(n - 1 for n in ns)\n",
-    "\n",
-    "    # Avoid division by zero\n",
-    "    if sd_denominator == 0:\n",
-    "        raise ValueError(\"Insufficient data to compute pooled standard deviation.\")\n",
-    "\n",
-    "    pooled_sample_sd = np.sqrt(sd_numerator / sd_denominator)\n",
-    "\n",
-    "    # Ensure pooled_sample_sd is not NaN or zero (to avoid division by zero later)\n",
-    "    if np.isnan(pooled_sample_sd) or pooled_sample_sd == 0:\n",
-    "        raise ValueError(\"Pooled sample standard deviation is NaN or zero.\")\n",
-    "\n",
-    "    out_delta_g, deltadelta = delta2_bootstrap_loop(x1, x2, x3, x4, resamples, pooled_sample_sd, random_seed, is_paired)\n",
-    "\n",
-    "    # Empirical delta_g calculation\n",
-    "    delta_g = ((np.mean(x4) - np.mean(x3)) - (np.mean(x2) - np.mean(x1))) / pooled_sample_sd\n",
+    "    \n",
+    "    if proportional:\n",
+    "        # For proportional data, pass 1.0 as dummy pooled_sd (won't be used)\n",
+    "        out_delta_g, deltadelta = delta2_bootstrap_loop(\n",
+    "            x1, x2, x3, x4, resamples, 1.0, random_seed, is_paired, proportional=True\n",
+    "        )\n",
+    "        # For proportional data, delta_g is the empirical delta-delta\n",
+    "        delta_g = ((np.mean(x4) - np.mean(x3)) - (np.mean(x2) - np.mean(x1)))\n",
+    "    else:\n",
+    "        # Calculate pooled sample standard deviation for non-proportional data\n",
+    "        stds = [np.std(x) for x in [x1, x2, x3, x4]]\n",
+    "        ns = [len(x) for x in [x1, x2, x3, x4]]\n",
+    "        \n",
+    "        sd_numerator = sum((n - 1) * s**2 for n, s in zip(ns, stds))\n",
+    "        sd_denominator = sum(n - 1 for n in ns)\n",
+    "        \n",
+    "        if sd_denominator == 0:\n",
+    "            raise ValueError(\"Insufficient data to compute pooled standard deviation.\")\n",
+    "            \n",
+    "        pooled_sample_sd = np.sqrt(sd_numerator / sd_denominator)\n",
+    "        \n",
+    "        if np.isnan(pooled_sample_sd) or pooled_sample_sd == 0:\n",
+    "            raise ValueError(\"Pooled sample standard deviation is NaN or zero.\")\n",
+    "            \n",
+    "        out_delta_g, deltadelta = delta2_bootstrap_loop(\n",
+    "            x1, x2, x3, x4, resamples, pooled_sample_sd, random_seed, is_paired, proportional=False\n",
+    "        )\n",
+    "        delta_g = ((np.mean(x4) - np.mean(x3)) - (np.mean(x2) - np.mean(x1))) / pooled_sample_sd\n",
     "\n",
     "    return out_delta_g, delta_g, deltadelta\n",
     "\n",

nbs/API/dabest_object.ipynb

@@ -64,6 +64,7 @@
    "source": [
     "#| export\n",
     "# Import standard data science libraries\n",
+    "import warnings\n",
     "from numpy import array, repeat, random, issubdtype, number\n",
     "import numpy as np\n",
     "import pandas as pd\n",
@@ -138,7 +139,6 @@
     "\n",
     "        # Check if there is NaN under any of the paired settings\n",
     "        if self.__is_paired and self.__output_data.isnull().values.any():\n",
-    "            import warnings\n",
     "            warn1 = f\"NaN values detected under paired setting and removed,\"\n",
     "            warn2 = f\" please check your data.\"\n",
     "            warnings.warn(warn1 + warn2)\n",
@@ -576,10 +576,10 @@
     "            if x is None:\n",
     "                error_msg = \"If `delta2` is True. `x` parameter cannot be None. String or list expected\"\n",
     "                raise ValueError(error_msg)\n",
-    "                \n",
+    "            \n",
     "            if self.__proportional:\n",
-    "                err0 = \"`proportional` and `delta2` cannot be True at the same time.\"\n",
-    "                raise ValueError(err0)\n",
+    "                mes1 = \"Only mean_diff is supported for proportional data when `delta2` is True\"\n",
+    "                warnings.warn(message=mes1, category=UserWarning)\n",
     "\n",
     "            # idx should not be specified\n",
     "            if idx:\n",
@@ -657,8 +657,6 @@
     "        \"\"\"\n",
     "        # Check if there is NaN under any of the paired settings\n",
     "        if self.__is_paired is not None and self.__output_data.isnull().values.any():\n",
-    "            print(\"Nan\")\n",
-    "            import warnings\n",
     "            warn1 = f\"NaN values detected under paired setting and removed,\"\n",
     "            warn2 = f\" please check your data.\"\n",
     "            warnings.warn(warn1 + warn2)\n",
@@ -710,7 +708,6 @@
     "\n",
     "            # Check if there is NaN under any of the paired settings\n",
     "            if self.__is_paired is not None and self.__output_data.isnull().values.any():\n",
-    "                import warnings\n",
     "                warn1 = f\"NaN values detected under paired setting and removed,\"\n",
     "                warn2 = f\" please check your data.\"\n",
     "                warnings.warn(warn1 + warn2)\n",

nbs/API/effsize_objects.ipynb

@@ -317,7 +317,8 @@
     "            raise ValueError(err1)\n",
     "\n",
     "        if self.__proportional and self.__effect_size not in [\"mean_diff\", \"cohens_h\"]:\n",
-    "            err1 = \"`proportional` is True; therefore effect size other than mean_diff and cohens_h is not defined.\"\n",
+    "            err1 = \"`proportional` is True; therefore effect size other than mean_diff and cohens_h is not defined.\" + \\\n",
+    "                    \"If you are calculating deltas' g, it's the same as delta-delta when `proportional` is True\"\n",
     "            raise ValueError(err1)\n",
     "\n",
     "        if self.__proportional and (\n",
@@ -1043,6 +1044,7 @@
     "                self.__is_paired,\n",
     "                self.__resamples,\n",
     "                self.__random_seed,\n",
+    "                self.__proportional,\n",
     "            )\n",
     "\n",
     "        for j, current_tuple in enumerate(idx):\n",

nbs/API/misc_tools.ipynb

@@ -643,12 +643,12 @@
     "        if color_by_subgroups:\n",
     "            plot_palette_raw = dict()\n",
     "            plot_palette_contrast = dict()\n",
-    "            # plot_palette_bar set to None because currently there is no empty_circle toggle for proportion plots\n",
-    "            plot_palette_bar = None\n",
+    "            plot_palette_bar = dict()\n",
     "            for i in range(len(idx)):\n",
     "                for names_i in idx[i]:\n",
     "                    plot_palette_raw[names_i] = swarm_colors[i]\n",
     "                    plot_palette_contrast[names_i] = contrast_colors[i]\n",
+    "                    plot_palette_bar[names_i] = bar_color[i]\n",
     "        else:\n",
     "            plot_palette_raw = dict(zip(categories, swarm_colors))\n",
     "            plot_palette_contrast = dict(zip(categories, contrast_colors))\n",
@@ -665,11 +665,12 @@
     "        if color_by_subgroups:\n",
     "            plot_palette_raw = dict()\n",
     "            plot_palette_contrast = dict()\n",
-    "            plot_palette_bar = None  # plot_palette_bar set to None because currently there is no empty_circle toggle for proportion plots\n",
+    "            plot_palette_bar = dict()\n",
     "            for i in range(len(idx)):\n",
     "                for names_i in idx[i]:\n",
     "                    plot_palette_raw[names_i] = swarm_colors[i]\n",
     "                    plot_palette_contrast[names_i] = contrast_colors[i]\n",
+    "                    plot_palette_bar[names_i] = bar_color[i]\n",
     "        else:\n",
     "            plot_palette_raw = dict(zip(names, swarm_colors))\n",
     "            plot_palette_contrast = dict(zip(names, contrast_colors))\n",
@@ -1071,6 +1072,7 @@
     "            ticks_with_counts.append(f\"{t}\\n(N={value})\")\n",
     "\n",
     "    fontsize_rawxlabel = plot_kwargs.get(\"fontsize_rawxlabel\")\n",
+    "    set_major_loc_method(plt.FixedLocator(get_ticks()))\n",
     "    set_label(ticks_with_counts, fontsize=fontsize_rawxlabel)\n",
     "\n",
     "    # Ensure ticks are at the correct locations\n",

nbs/API/plot_tools.ipynb

@@ -781,14 +781,17 @@
     "    right_idx = []\n",
     "    # Design for Sankey Flow Diagram\n",
     "    sankey_idx = (\n",
-    "        [\n",
-    "            (control, test)\n",
-    "            for i in idx\n",
-    "            for control, test in zip(i[:], (i[1:] + (i[0],)))\n",
-    "        ]\n",
-    "        if flow\n",
-    "        else temp_idx\n",
-    "    )\n",
+    "    [\n",
+    "        (control, test)\n",
+    "        for i in idx\n",
+    "        for control, test in zip(\n",
+    "            i[:],\n",
+    "            (tuple(i[1:]) + (i[0],)) if isinstance(i, tuple) else (list(i[1:]) + [i[0]])\n",
+    "        )\n",
+    "    ]\n",
+    "    if flow\n",
+    "    else temp_idx\n",
+    ")\n",
     "    for i in sankey_idx:\n",
     "        left_idx.append(i[0])\n",
     "        right_idx.append(i[1])\n",
@@ -2115,6 +2118,7 @@
     "        plot_data: pd.DataFrame, \n",
     "        bar_color: str, \n",
     "        plot_palette_bar: dict, \n",
+    "        color_col: str,\n",
     "        plot_kwargs: dict, \n",
     "        barplot_kwargs: dict, \n",
     "        horizontal: bool\n",
@@ -2138,6 +2142,8 @@
     "        Color of the bar.\n",
     "    plot_palette_bar : dict\n",
     "        Dictionary of colors used in the bar plot.\n",
+    "    color_col : str\n",
+    "        Column name of the color column.\n",
     "    plot_kwargs : dict\n",
     "        Keyword arguments for the plot.\n",
     "    barplot_kwargs : dict\n",
@@ -2152,7 +2158,26 @@
     "    else:\n",
     "        x_var, y_var, orient = all_plot_groups, np.ones(len(all_plot_groups)), \"v\"\n",
     "\n",
-    "    bar1_df = pd.DataFrame({xvar: x_var, \"proportion\": y_var})\n",
+    "        # Create bar1_df with basic columns\n",
+    "    bar1_df = pd.DataFrame({\n",
+    "        xvar: x_var, \n",
+    "        \"proportion\": y_var\n",
+    "    })\n",
+    "\n",
+    "    # Handle colors\n",
+    "    if color_col:\n",
+    "        # Get first color value for each group\n",
+    "        color_mapping = plot_data.groupby(xvar, observed=False)[color_col].first()\n",
+    "        bar1_df[color_col] = [color_mapping.get(group) for group in all_plot_groups]\n",
+    "        \n",
+    "        # Map colors, defaulting to bar_color if no match\n",
+    "        edge_colors = [\n",
+    "            plot_palette_bar.get(hue_val, bar_color) \n",
+    "            for hue_val in bar1_df[color_col]\n",
+    "        ]\n",
+    "    else:\n",
+    "        edge_colors = bar_color\n",
+    "\n",
     "\n",
     "    bar1 = sns.barplot(\n",
     "        data=bar1_df,\n",
@@ -2162,7 +2187,7 @@
     "        order=all_plot_groups,\n",
     "        linewidth=2,\n",
     "        facecolor=(1, 1, 1, 0),\n",
-    "        edgecolor=bar_color,\n",
+    "        edgecolor=edge_colors,\n",
     "        zorder=1,\n",
     "        orient=orient,\n",
     "    )\n",
@@ -2173,6 +2198,8 @@
     "        ax=rawdata_axes,\n",
     "        order=all_plot_groups,\n",
     "        palette=plot_palette_bar,\n",
+    "        hue=color_col,\n",
+    "        dodge=False,\n",
     "        zorder=1,\n",
     "        orient=orient,\n",
     "        **barplot_kwargs\n",

nbs/API/plotter.ipynb

@@ -334,6 +334,7 @@
     "                plot_data = plot_data, \n",
     "                bar_color = bar_color, \n",
     "                plot_palette_bar = plot_palette_bar, \n",
+    "                color_col = color_col,\n",
     "                plot_kwargs = plot_kwargs, \n",
     "                barplot_kwargs = barplot_kwargs,\n",
     "                horizontal = horizontal,\n",