Rearrange and clarify

jswelling · Feb 1, 2024 · 9307900 · 9307900
1 parent cc48bd8
commit 9307900
Showing 1 changed file with 51 additions and 21 deletions.
diff --git a/notebooks/seaborn_workbook_3.ipynb b/notebooks/seaborn_workbook_3.ipynb
@@ -243,7 +243,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "c30ec428",
+   "id": "0695faeb",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -428,7 +428,7 @@
   },
   {
    "cell_type": "markdown",
-   "id": "760523e6",
+   "id": "a2dbd581",
    "metadata": {},
    "source": [
     "## Bootstrapped Confidence Intervals vs. Standard Deviation\n",
@@ -439,7 +439,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "f55a8865",
+   "id": "4ec45b94",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -448,7 +448,7 @@
   },
   {
    "cell_type": "markdown",
-   "id": "ccf01e46",
+   "id": "c0adf436",
    "metadata": {},
    "source": [
     "### Turn off the error bars entirely"
@@ -457,7 +457,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "66f7afeb",
+   "id": "88e19a82",
    "metadata": {
     "scrolled": false
    },
@@ -468,82 +468,112 @@
   },
   {
    "cell_type": "markdown",
-   "id": "0676df46",
+   "id": "4e068013",
    "metadata": {},
    "source": [
-    "### *percentile interval* (non-parametric)"
+    "### *standard deviation* (parametric)\n",
+    "1 standard deviation on each side of the mean by default, so for normal data 68% will fall within the range."
    ]
   },
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "d53750a7",
+   "id": "3180b969",
    "metadata": {
     "scrolled": false
    },
    "outputs": [],
    "source": [
-    "sns.relplot(x=\"timepoint\", y=\"signal\", kind=\"line\", data=fmri_df, col=\"region\", hue=\"event\", errorbar=\"pi\");"
+    "sns.relplot(x=\"timepoint\", y=\"signal\", kind=\"line\", data=fmri_df, col=\"region\", hue=\"event\", errorbar=\"sd\");"
    ]
   },
   {
    "cell_type": "markdown",
-   "id": "6d81bce6",
+   "id": "c0944dae",
    "metadata": {},
    "source": [
-    "### *confidence interval* (non-parametric)"
+    "### *standard error* (parametric)\n",
+    "How far is the estimated mean likely to be from the real mean?\n",
+    "* This one is just standard deviation over sqrt(number of samples)\n",
+    "* Thus, more samples makes the standard error smaller"
    ]
   },
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "fc19f6de",
+   "id": "cd89f0fc",
    "metadata": {
     "scrolled": false
    },
    "outputs": [],
    "source": [
-    "sns.relplot(x=\"timepoint\", y=\"signal\", kind=\"line\", data=fmri_df, col=\"region\", hue=\"event\", errorbar=\"ci\");"
+    "sns.relplot(x=\"timepoint\", y=\"signal\", kind=\"line\", data=fmri_df, col=\"region\", hue=\"event\", errorbar=\"se\");"
    ]
   },
   {
    "cell_type": "markdown",
-   "id": "dac75ce9",
+   "id": "f7ed22f7",
    "metadata": {},
    "source": [
-    "### *standard deviation* (parametric)"
+    "### *percentile interval* (non-parametric)\n",
+    "How confident are we that a given sample will fall in the range?\n",
+    "* 95% confidence by default - note that it is wider than the 'standard deviation' bars\n",
+    "* Note also that it can be assymmetric, unlike the parametric statistics."
    ]
   },
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "35d7435d",
+   "id": "694e6a7b",
    "metadata": {
     "scrolled": false
    },
    "outputs": [],
    "source": [
-    "sns.relplot(x=\"timepoint\", y=\"signal\", kind=\"line\", data=fmri_df, col=\"region\", hue=\"event\", errorbar=\"sd\");"
+    "sns.relplot(x=\"timepoint\", y=\"signal\", kind=\"line\", data=fmri_df, col=\"region\", hue=\"event\", errorbar=\"pi\");"
    ]
   },
   {
    "cell_type": "markdown",
-   "id": "fbd9737b",
+   "id": "3ed82706",
    "metadata": {},
    "source": [
-    "### *standard error* (parametric)"
+    "### Repeat that with a 50% confidence interval.\n",
+    "This is equivalent to the Q1/median/Q3 graphs we were drawing earlier."
    ]
   },
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "be4c2ea9",
+   "id": "59acff5d",
    "metadata": {
     "scrolled": false
    },
    "outputs": [],
    "source": [
-    "sns.relplot(x=\"timepoint\", y=\"signal\", kind=\"line\", data=fmri_df, col=\"region\", hue=\"event\", errorbar=\"se\");"
+    "sns.relplot(x=\"timepoint\", y=\"signal\", kind=\"line\", data=fmri_df, col=\"region\", hue=\"event\", errorbar=(\"pi\",50));"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "8fc54cb8",
+   "metadata": {},
+   "source": [
+    "### *confidence interval* (non-parametric)\n",
+    "These are computed by bootstrapping, which takes a bit of time.\n",
+    "What we see is the percentile interval of the bootstrap distribution."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "61789053",
+   "metadata": {
+    "scrolled": false
+   },
+   "outputs": [],
+   "source": [
+    "sns.relplot(x=\"timepoint\", y=\"signal\", kind=\"line\", data=fmri_df, col=\"region\", hue=\"event\", errorbar=\"ci\");"
    ]
   },
   {