diff --git a/lecture15.ipynb b/lecture15.ipynb index 96d96bd57147a860b35412dd4bc25781b3a1bef7..fce4ce6689b735d4c631863701a03da3792bc02d 100644 --- a/lecture15.ipynb +++ b/lecture15.ipynb @@ -2,7 +2,21 @@ "cells": [ { "cell_type": "markdown", - "metadata": {}, + "metadata": { + "id": "01316CBFA318429FA9F64C5E67A0DFED", + "jupyter": {}, + "notebookId": "676d00454d522334965beb55", + "runtime": { + "execution_status": null, + "is_visible": false, + "status": "default" + }, + "scrolled": false, + "slideshow": { + "slide_type": "slide" + }, + "tags": [] + }, "source": [ "#
Lecture15 : 课程回顾和复习
\n", " \n", @@ -11,123 +25,249 @@ }, { "cell_type": "markdown", - "metadata": {}, + "metadata": { + "id": "73B298DFD1F94567962C5C022C3C1921", + "jupyter": {}, + "notebookId": "676d00454d522334965beb55", + "runtime": { + "execution_status": null, + "is_visible": false, + "status": "default" + }, + "scrolled": false, + "slideshow": { + "slide_type": "slide" + }, + "tags": [] + }, "source": [ - "## Outlines\n", + "## Outlines \n", "\n", - "| 序号 | 课程内容 |\n", - "| :--: | :--------------------------------------------: |\n", + "| 序号 | 课程内容 | \n", + "| :--: | :--------------------------------------------: | \n", "| 1 | 课程介绍 | \n", - "| 2 | Bayes' Rule |\n", - "| 3 | The Beta-Binomial Bayesian Model |\n", - "| 4 | Balance and Sequentiality in Bayesian Analyses |\n", - "| 5 | Approximating the Posterior |\n", - "| 6 | MCMC under the Hood |\n", - "| 7 | Posterior Inference & Prediction |\n", - "| 8 | A Simple Normal Regression |\n", - "| 9 | Bayes factors |\n", - "| 10 | Multiple regression |\n", - "| 11 | Evaluating Regression Models |\n", - "| 12 | GLM: Logistic Regression |\n", - "| 13 | Hierarchical Models 1 |\n", + "| 2 | Bayes' Rule | \n", + "| 3 | The Beta-Binomial Bayesian Model | \n", + "| 4 | Balance and Sequentiality in Bayesian Analyses | \n", + "| 5 | Approximating the Posterior | \n", + "| 6 | MCMC under the Hood | \n", + "| 7 | Posterior Inference & Prediction | \n", + "| 8 | A Simple Normal Regression | \n", + "| 9 | Bayes factors | \n", + "| 10 | Multiple regression | \n", + "| 11 | Evaluating Regression Models | \n", + "| 12 | GLM: Logistic Regression | \n", + "| 13 | Hierarchical Models 1 | \n", "| 14 | Hierarchical Models 2 |" ] }, { "cell_type": "markdown", - "metadata": {}, + "metadata": { + "id": "C1E28127BE084E798D2E60EAA5CC8AC1", + "jupyter": {}, + "notebookId": "676d00454d522334965beb55", + "runtime": { + "execution_status": null, + "is_visible": false, + "status": "default" + }, + "scrolled": false, + "slideshow": { + "slide_type": "slide" + }, + "tags": [] + }, "source": [ - "## 对本课程最简略的概括\n", + "## 对本课程最简略的概括 \n", "\n", - "- 一个概念：贝叶斯视角下的参数\n", - "- 一个公式：贝叶斯公式\n", - "- 一个算法：马尔科夫链蒙特卡洛（MCMC）\n", - "- 一个软件：PyMC\n", - "- 一个workflow：贝叶斯分析的工作流程\n" + "- 一个概念：贝叶斯视角下的参数 \n", + "- 一个公式：贝叶斯公式 \n", + "- 一个算法：马尔科夫链蒙特卡洛（MCMC） \n", + "- 一个软件：PyMC \n", + "- 一个workflow：贝叶斯分析的工作流程 \n" ] }, { "cell_type": "markdown", - "metadata": {}, + "metadata": { + "id": "EFF8F175A46C4854849AE68D209C4134", + "jupyter": {}, + "notebookId": "676d00454d522334965beb55", + "runtime": { + "execution_status": null, + "is_visible": false, + "status": "default" + }, + "scrolled": false, + "slideshow": { + "slide_type": "slide" + }, + "tags": [] + }, "source": [ - "## 一个概念\n", - "**模型参数是随机的，是概率分布**\n", + "## 一个概念 \n", + "**模型参数是随机的，是概率分布** \n", "\n", "\n" ] }, { "cell_type": "markdown", - "metadata": {}, + "metadata": { + "id": "D66BD46E254849FBB15781DED8571C9B", + "jupyter": {}, + "notebookId": "676d00454d522334965beb55", + "runtime": { + "execution_status": null, + "is_visible": false, + "status": "default" + }, + "scrolled": false, + "slideshow": { + "slide_type": "slide" + }, + "tags": [] + }, "source": [ "\n", - "## 一个公式\n", + "## 一个公式 \n", "\n", "$$ \n", "P(A|B) = \\frac{P(A) * P (B | A)}{P(B)} \n", "$$ \n", "\n", - "- $P(A|B)$: 后验概率\n", - "- $P(A)$: 先验概率\n", - "- $P(B|A)$: 似然函数\n", - "- $P(B)$: Marginal likelihood\n", + "- $P(A|B)$: 后验概率 \n", + "- $P(A)$: 先验概率 \n", + "- $P(B|A)$: 似然函数 \n", + "- $P(B)$: Marginal likelihood \n", "\n", "
" ] }, { "cell_type": "markdown", - "metadata": {}, + "metadata": { + "id": "02CC2561821146098794C9A6623D666D", + "jupyter": {}, + "notebookId": "676d00454d522334965beb55", + "runtime": { + "execution_status": null, + "is_visible": false, + "status": "default" + }, + "scrolled": false, + "slideshow": { + "slide_type": "slide" + }, + "tags": [] + }, "source": [ - "## 一个算法\n", + "## 一个算法 \n", "马尔科夫链蒙特卡洛（MCMC）" ] }, { "cell_type": "markdown", - "metadata": {}, + "metadata": { + "id": "7FD5E91C27D44460AB9E5D3E707383C3", + "jupyter": {}, + "notebookId": "676d00454d522334965beb55", + "runtime": { + "execution_status": null, + "is_visible": false, + "status": "default" + }, + "scrolled": false, + "slideshow": { + "slide_type": "slide" + }, + "tags": [] + }, "source": [ - "## 一个软件包\n", - "**PyMC**\n", + "## 一个软件包 \n", + "**PyMC** \n", "\n", - "通过PyMC学习使用概率编程语言(probability programming language)，实现贝叶斯推断。\n", + "通过PyMC学习使用概率编程语言(probability programming language)，实现贝叶斯推断。 \n", "\n", "![Image Name](https://cdn.kesci.com/upload/sl1bdlkzgo.png?imageView2/0/w/640/h/640) \n" ] }, { "cell_type": "markdown", - "metadata": {}, + "metadata": { + "id": "931A1760068B43F284ECF2FEB63FA832", + "jupyter": {}, + "notebookId": "676d00454d522334965beb55", + "runtime": { + "execution_status": null, + "is_visible": false, + "status": "default" + }, + "scrolled": false, + "slideshow": { + "slide_type": "slide" + }, + "tags": [] + }, "source": [ - "## 一个数据分析流程\n", + "## 一个数据分析流程 \n", "\n", - "**Bayesian Workflow**\n", + "**Bayesian Workflow** \n", "\n", - "![Image Name](https://cdn.kesci.com/upload/sozk5mh1vf.png?imageView2/0/w/960/h/960)\n", + "![Image Name](https://cdn.kesci.com/upload/sozk5mh1vf.png?imageView2/0/w/960/h/960) \n", "\n", - "其他相关指南参考：\n", + "其他相关指南参考： \n", "\n", - "> Kruschke, J.K. Bayesian Analysis Reporting Guidelines. Nat Hum Behav 5, 1282–1291 (2021). https://doi.org/10.1038/s41562-021-01177-7\n" + "> Kruschke, J.K. Bayesian Analysis Reporting Guidelines. Nat Hum Behav 5, 1282–1291 (2021). https://doi.org/10.1038/s41562-021-01177-7 \n" ] }, { "cell_type": "markdown", - "metadata": {}, + "metadata": { + "id": "D7B35CD9BC70483EB06EEDE20FF4431A", + "jupyter": {}, + "notebookId": "676d00454d522334965beb55", + "runtime": { + "execution_status": null, + "is_visible": false, + "status": "default" + }, + "scrolled": false, + "slideshow": { + "slide_type": "slide" + }, + "tags": [] + }, "source": [ - "## 一个完整的例子\n", + "## 一个完整的例子 \n", "\n", - "### 研究问题\n", + "### 研究问题 \n", "\n", "“**随机点运动范式中，反应时间如何受到随机点运动方向一致性比例的影响，如果会的话，其影响程度是怎么样的**？”" ] }, { "cell_type": "markdown", - "metadata": {}, + "metadata": { + "id": "C3DB1AAEE037490EB2DE47764B5EBC1D", + "jupyter": {}, + "notebookId": "676d00454d522334965beb55", + "runtime": { + "execution_status": null, + "is_visible": false, + "status": "default" + }, + "scrolled": false, + "slideshow": { + "slide_type": "slide" + }, + "tags": [] + }, "source": [ - "### 数据\n", + "### 数据 \n", "\n", - "Evans et al.（2020, Exp. 1） 的数据，包括57名被试的数据，单因素被试内实验设计，自变量为2个水平。\n", + "Evans et al.（2020, Exp. 1） 的数据，包括57名被试的数据，单因素被试内实验设计，自变量为2个水平。 \n", "\n", "
\n", " \n", @@ -148,16 +288,19 @@ { "cell_type": "code", "execution_count": 1, - "metadata": {}, - "outputs": [ - { - "name": "stderr", - "output_type": "stream", - "text": [ - "WARNING (pytensor.tensor.blas): Using NumPy C-API based implementation for BLAS functions.\n" - ] - } - ], + "metadata": { + "collapsed": false, + "id": "3A40117637B64A5EACEF450E81A5F744", + "jupyter": {}, + "notebookId": "676d00454d522334965beb55", + "scrolled": false, + "slideshow": { + "slide_type": "slide" + }, + "tags": [], + "trusted": true + }, + "outputs": [], "source": [ "# 导入 pymc 模型包，和 arviz 等分析工具 \n", "import pymc as pm\n", @@ -178,8 +321,19 @@ }, { "cell_type": "code", - "execution_count": 35, - "metadata": {}, + "execution_count": 2, + "metadata": { + "collapsed": false, + "id": "21D2D49641994076BFEFFCC702C1AD13", + "jupyter": {}, + "notebookId": "676d00454d522334965beb55", + "scrolled": false, + "slideshow": { + "slide_type": "slide" + }, + "tags": [], + "trusted": true + }, "outputs": [ { "data": { @@ -223,103 +377,103 @@ " \n", " \n", " \n", - " \n", - " \n", - " \n", - " \n", + " \n", + " \n", " \n", + " \n", + " \n", " \n", - " \n", - " \n", - " \n", + " \n", + " \n", + " \n", " \n", - " \n", - " \n", - " \n", - " \n", - " \n", + " \n", + " \n", + " \n", + " \n", + " \n", " \n", - " \n", + " \n", " \n", " \n", " \n", - " \n", - " \n", + " \n", + " \n", " \n", " \n", " \n", " \n", " \n", " \n", - " \n", - " \n", - " \n", - " \n", - " \n", + " \n", " \n", - " \n", + " \n", + " \n", + " \n", + " \n", + " \n", " \n", - " \n", + " \n", " \n", " \n", " \n", - " \n", - " \n", + " \n", + " \n", " \n", - " \n", + " \n", " \n", " \n", " \n", " \n", " \n", " \n", - " \n", - " \n", - " \n", + " \n", + " \n", + " \n", " \n", - " \n", + " \n", " \n", - " \n", + " \n", " \n", " \n", " \n", - " \n", - " \n", - " \n", - " \n", + " \n", + " \n", " \n", + " \n", + " \n", " \n", - " \n", - " \n", - " \n", + " \n", + " \n", + " \n", " \n", - " \n", - " \n", - " \n", - " \n", - " \n", + " \n", + " \n", + " \n", + " \n", + " \n", " \n", - " \n", + " \n", " \n", " \n", " \n", - " \n", - " \n", + " \n", + " \n", " \n", - " \n", + " \n", " \n", " \n", " \n", - " \n", - " \n", + " \n", + " \n", " \n", " \n", - " \n", - " \n", + " \n", + " \n", " \n", - " \n", + " \n", " \n", - " \n", + " \n", " \n", " \n", " \n", @@ -327,29 +481,29 @@ "" ], "text/plain": [ - " subject blkNum trlNum coherentDots numberofDots percentCoherence \\\n", - "2239 66670 2 2 2 40 5 \n", - "2240 66670 2 3 2 40 5 \n", - "2241 66670 2 4 4 40 10 \n", - "2243 66670 2 6 2 40 5 \n", - "2244 66670 2 7 4 40 10 \n", - "\n", - " winningDirection response correct eventCount averageFrameRate RT \\\n", - "2239 left left 1 23 15.700 1465 \n", - "2240 left left 1 25 15.375 1626 \n", - "2241 right right 1 37 15.346 2411 \n", - "2243 left left 1 16 16.113 993 \n", - "2244 left left 1 16 15.564 1028 \n", - "\n", - " Coherence subj_id obs_id log_RTs global_id \n", - "2239 0 66670 1 7.289611 0 \n", - "2240 0 66670 2 7.393878 1 \n", - "2241 1 66670 3 7.787797 2 \n", - "2243 0 66670 4 6.900731 3 \n", - "2244 1 66670 5 6.935370 4 " + " subject blkNum trlNum coherentDots numberofDots percentCoherence \\\n", + "0 31727 2 1 4 40 10 \n", + "2 31727 2 3 2 40 5 \n", + "6 31727 2 7 4 40 10 \n", + "7 31727 2 8 4 40 10 \n", + "9 31727 2 10 4 40 10 \n", + "\n", + " winningDirection response correct eventCount averageFrameRate RT \\\n", + "0 right right 1 51 14.452 3529 \n", + "2 left right 0 15 15.322 979 \n", + "6 right right 1 48 14.307 3355 \n", + "7 right right 1 31 14.472 2142 \n", + "9 right right 1 16 15.166 1055 \n", + "\n", + " Coherence subj_id obs_id log_RTs global_id \n", + "0 1 31727 1 8.168770 0 \n", + "2 0 31727 2 6.886532 1 \n", + "6 1 31727 3 8.118207 2 \n", + "7 1 31727 4 7.669495 3 \n", + "9 1 31727 5 6.961296 4 " ] }, - "execution_count": 35, + "execution_count": 2, "metadata": {}, "output_type": "execute_result" } @@ -360,10 +514,18 @@ " df_raw = pd.read_csv(\"/home/mw/input/bayes3797/evans2020JExpPsycholLearn_exp1_full_data.csv\") \n", "except:\n", " df_raw = pd.read_csv('data/evans2020JExpPsycholLearn_exp1_full_data.csv')\n", + "\n", + "# 筛选出特定被试并创建索引\n", + "df = df_raw.query(\"percentCoherence in [5, 10]\").copy()\n", + "df[\"Coherence\"] = np.where(df[\"percentCoherence\"] == 5, 0, 1)\n", + "\n", + "# 随机选择30个被试\n", + "selected_subjects = df['subject'].drop_duplicates().sample(n=30, random_state=42)\n", + "df = df[df['subject'].isin(selected_subjects)]\n", + "\n", "# 为每个被试建立索引 'subj_id' 和 'obs_id'\n", - "df = df_raw.copy()\n", - "df['subj_id'] = df_raw['subject']\n", - "df['obs_id'] = df_raw.groupby('subject').cumcount() + 1\n", + "df['subj_id'] = df['subject']\n", + "df['obs_id'] = df.groupby('subject').cumcount() + 1\n", "\n", "# 对反应时间取对数\n", "df[\"log_RTs\"] = np.log(df[\"RT\"])\n", @@ -376,37 +538,51 @@ }, { "cell_type": "markdown", - "metadata": {}, + "metadata": { + "id": "E46E0DA27BE04AFA96EA526FEB6632C1", + "jupyter": {}, + "notebookId": "676d00454d522334965beb55", + "runtime": { + "execution_status": null, + "is_visible": false, + "status": "default" + }, + "scrolled": false, + "slideshow": { + "slide_type": "slide" + }, + "tags": [] + }, "source": [ - "### 模型设定\n", + "### 模型设定 \n", "\n", - "建立三个不同的模型来探讨反应时间与一致性比例之间的关系。\n", + "建立三个不同的模型来探讨反应时间与一致性比例之间的关系。 \n", "\n", - "#### 模型1：完全池化模型\n", + "#### 模型1：完全池化模型 \n", "\n", - "**目的：不考虑随机点运动方向一致性的比例对反应时间的影响**\n", + "**目的：不考虑随机点运动方向一致性的比例对反应时间的影响** \n", "\n", "\n", "$$ \n", "\\begin{array}{lcrl} \n", "Y_i | \\beta_0, \\beta_1, \\sigma & \\stackrel{ind}{\\sim} N\\left(\\mu_i, \\sigma^2\\right) \\;\\; \\text{ with } \\;\\; \\mu_i = \\beta_0 + \\beta_1X_i \\\\ \n", "\\end{array} \n", - "$$ \n", + "$$ \n", "\n", "\n", - "#### 模型2：部分池化模型（变化截距）\n", + "#### 模型2：部分池化模型（变化截距） \n", "\n", - "**目的：随机点运动方向的一致性与反应时间之间的关系在被试内有什么不同**\n", + "**目的：随机点运动方向的一致性与反应时间之间的关系在被试内有什么不同** \n", "\n", "$$ \n", - "\\begin{array}{rll}\n", + "\\begin{array}{rll} \n", "&\\mu_{\\beta_0}, \\sigma_{\\beta_0} & \\text{Layer 3: 总体水平} \\\\ \n", "\\beta_{0j} | \\mu_{\\beta_0}, \\sigma_{\\beta_0} & \\stackrel{\\text{ind}}{\\sim} N(\\mu_{\\beta_0}, \\sigma_{\\beta_0}^2) & \\text{Layer 2: 组水平} \\\\ \n", "Y_{ij} | \\beta_{0j}, \\beta_{1}, \\sigma_y & \\sim N(\\mu_{ij}, \\sigma_y^2) \\;\\; \\text{ with } \\;\\; \\mu_{ij} = \\beta_{0j} + \\beta_{1} X_{ij} & \\text{Layer 1: 试次水平/数据点} \\\\ \n", "\\end{array} \n", - "$$\n", + "$$ \n", "\n", - "#### 模型3：部分池化模型（变化斜率和截距）\n", + "#### 模型3：部分池化模型（变化斜率和截距） \n", "\n", "**目的：考虑截距和斜率共同变化的情况，并全局参数进行定义，即** \n", "\n", @@ -425,11 +601,18 @@ }, { "cell_type": "code", - "execution_count": 76, + "execution_count": 3, "metadata": { + "collapsed": false, + "id": "7186B62EB8094A52A5B9B471D4A1CBCC", + "jupyter": {}, + "notebookId": "676d00454d522334965beb55", + "scrolled": false, "slideshow": { "slide_type": "fragment" - } + }, + "tags": [], + "trusted": true }, "outputs": [], "source": [ @@ -439,11 +622,18 @@ }, { "cell_type": "code", - "execution_count": 77, + "execution_count": 4, "metadata": { + "collapsed": false, + "id": "F9F17B7DB7BC48EBAA73EEC08749AC03", + "jupyter": {}, + "notebookId": "676d00454d522334965beb55", + "scrolled": false, "slideshow": { "slide_type": "fragment" - } + }, + "tags": [], + "trusted": true }, "outputs": [], "source": [ @@ -453,54 +643,115 @@ }, { "cell_type": "code", - "execution_count": 78, - "metadata": {}, + "execution_count": 3, + "metadata": { + "collapsed": false, + "id": "240B1895D6DA4CDD9781DE016DB92970", + "jupyter": {}, + "notebookId": "676d00454d522334965beb55", + "scrolled": false, + "slideshow": { + "slide_type": "slide" + }, + "tags": [], + "trusted": true + }, "outputs": [], "source": [ "# 模型3：随机截距和斜率模型 \n", "var_both_model = bmb.Model(\"log_RTs ~ Coherence + (Coherence|subj_id)\",df)" ] }, + { + "cell_type": "code", + "execution_count": 4, + "metadata": { + "collapsed": false, + "id": "D5D88F9CCA504CBAA4A84E43DAA2F591", + "jupyter": {}, + "notebookId": "676d00454d522334965beb55", + "scrolled": false, + "slideshow": { + "slide_type": "slide" + }, + "tags": [], + "trusted": true + }, + "outputs": [], + "source": [ + "priors = {\"Coherence\": bmb.Prior(\"Normal\", mu=0, sigma=0.2)}\n", + "\n", + "var_both_model = bmb.Model(\"log_RTs ~ Coherence + (Coherence|subj_id)\",df, priors=priors)" + ] + }, { "cell_type": "markdown", - "metadata": {}, + "metadata": { + "id": "C08AEA6F611B4BD59B362D55FC720CEC", + "jupyter": {}, + "notebookId": "676d00454d522334965beb55", + "runtime": { + "execution_status": null, + "is_visible": false, + "status": "default" + }, + "scrolled": false, + "slideshow": { + "slide_type": "slide" + }, + "tags": [] + }, "source": [ - "### 拟合数据及MCMC评估\n", + "### 拟合数据及MCMC评估 \n", "\n", "接下来会对3个模型进行数据拟合、MCMC评估及后验计算。" ] }, { "cell_type": "markdown", - "metadata": {}, + "metadata": { + "id": "CED2FC0F980C44EEA742091827F358D0", + "jupyter": {}, + "notebookId": "676d00454d522334965beb55", + "runtime": { + "execution_status": null, + "is_visible": false, + "status": "default" + }, + "scrolled": false, + "slideshow": { + "slide_type": "slide" + }, + "tags": [] + }, "source": [ "**模型1**：完全池化模型" ] }, { "cell_type": "code", - "execution_count": 79, - "metadata": {}, - "outputs": [ - { - "name": "stderr", - "output_type": "stream", - "text": [ - "Auto-assigning NUTS sampler...\n", - "Initializing NUTS using jitter+adapt_diag...\n", - "Multiprocess sampling (4 chains in 4 jobs)\n", - "NUTS: [log_RTs_sigma, Intercept, Coherence]\n" - ] + "execution_count": 8, + "metadata": { + "collapsed": false, + "id": "7189093C10AF4734A3BFE73E7CF4C90C", + "jupyter": {}, + "notebookId": "676d00454d522334965beb55", + "scrolled": false, + "slideshow": { + "slide_type": "slide" }, + "tags": [], + "trusted": true + }, + "outputs": [ { "data": { - "application/vnd.jupyter.widget-view+json": { - "model_id": "3b2e2a7e57be436aa52a55b0cba81da9", - "version_major": 2, - "version_minor": 0 - }, + "text/html": [ + "
Sampling 4 chains, 0 divergences ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 100% 0:00:00 / 0:00:04\n",
+       "
\n" + ], "text/plain": [ - "Output()" + "Sampling 4 chains, 0 divergences \u001b[32m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[35m100%\u001b[0m \u001b[36m0:00:00\u001b[0m / \u001b[33m0:00:04\u001b[0m\n" ] }, "metadata": {}, @@ -533,7 +784,7 @@ "name": "stderr", "output_type": "stream", "text": [ - "Sampling 4 chains for 1_000 tune and 1_000 draw iterations (4_000 + 4_000 draws total) took 20 seconds.\n" + "Sampling 4 chains for 1_000 tune and 1_000 draw iterations (4_000 + 4_000 draws total) took 5 seconds.\n" ] } ], @@ -543,8 +794,19 @@ }, { "cell_type": "code", - "execution_count": 47, - "metadata": {}, + "execution_count": 13, + "metadata": { + "collapsed": false, + "id": "582852B4FF454FD494062A437D9BEAC2", + "jupyter": {}, + "notebookId": "676d00454d522334965beb55", + "scrolled": false, + "slideshow": { + "slide_type": "slide" + }, + "tags": [], + "trusted": true + }, "outputs": [ { "data": { @@ -580,40 +842,40 @@ " \n", " \n", " \n", - " \n", - " \n", - " \n", - " \n", - " \n", + " \n", + " \n", + " \n", + " \n", + " \n", " \n", " \n", - " \n", - " \n", - " \n", + " \n", + " \n", + " \n", " \n", " \n", - " \n", - " \n", - " \n", - " \n", - " \n", + " \n", + " \n", + " \n", + " \n", + " \n", " \n", " \n", - " \n", - " \n", - " \n", + " \n", + " \n", + " \n", " \n", " \n", " \n", - " \n", - " \n", - " \n", - " \n", + " \n", + " \n", + " \n", + " \n", " \n", " \n", - " \n", - " \n", - " \n", + " \n", + " \n", + " \n", " \n", " \n", "
22396667022031727214405leftleft10rightright12315.70014650666705114.452352913172717.2896118.1687700
224066670231727232405leftleft12515.3751626right0666701515.32297903172727.3938786.8865321
22416667063172724744010rightright13715.34624114814.30733551666703172737.7877978.1182072
22436667026731727284405leftleft10rightright11616.1139930666703114.472214213172746.9007317.6694953
224466670931727271044010leftleftrightright11615.564102815.16610551666703172756.9353706.9612964
Intercept6.9510.0216.9146.992Coherence-0.1060.009-0.123-0.0900.00.05789.02878.01.06412.03452.01.01
Coherence-0.0570.029-0.115-0.006Intercept6.9190.0066.9076.9300.00.05754.03162.01.06641.03499.01.00
log_RTs_sigma0.7090.0100.6900.7290.5630.0030.5570.5690.00.05492.03050.01.05618.02994.01.00
\n", @@ -621,17 +883,17 @@ ], "text/plain": [ " mean sd hdi_3% hdi_97% mcse_mean mcse_sd ess_bulk \\\n", - "Intercept 6.951 0.021 6.914 6.992 0.0 0.0 5789.0 \n", - "Coherence -0.057 0.029 -0.115 -0.006 0.0 0.0 5754.0 \n", - "log_RTs_sigma 0.709 0.010 0.690 0.729 0.0 0.0 5492.0 \n", + "Coherence -0.106 0.009 -0.123 -0.090 0.0 0.0 6412.0 \n", + "Intercept 6.919 0.006 6.907 6.930 0.0 0.0 6641.0 \n", + "log_RTs_sigma 0.563 0.003 0.557 0.569 0.0 0.0 5618.0 \n", "\n", " ess_tail r_hat \n", - "Intercept 2878.0 1.0 \n", - "Coherence 3162.0 1.0 \n", - "log_RTs_sigma 3050.0 1.0 " + "Coherence 3452.0 1.01 \n", + "Intercept 3499.0 1.00 \n", + "log_RTs_sigma 2994.0 1.00 " ] }, - "execution_count": 47, + "execution_count": 13, "metadata": {}, "output_type": "execute_result" } @@ -643,22 +905,35 @@ }, { "cell_type": "code", - "execution_count": 48, - "metadata": {}, + "execution_count": 14, + "metadata": { + "collapsed": false, + "id": "ADBD5B12CD7D48E990494E922E74DD6C", + "jupyter": {}, + "notebookId": "676d00454d522334965beb55", + "scrolled": false, + "slideshow": { + "slide_type": "slide" + }, + "tags": [], + "trusted": true + }, "outputs": [ { "data": { "text/plain": [ - "array([], dtype=object)" + "array([], dtype=object)" ] }, - "execution_count": 48, + "execution_count": 14, "metadata": {}, "output_type": "execute_result" }, { "data": { - "image/png": "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", + "text/html": [ + "" + ], "text/plain": [ "
" ] @@ -677,35 +952,49 @@ }, { "cell_type": "markdown", - "metadata": {}, + "metadata": { + "id": "F3C5AEE4905443EB9E6FC50380ACA4DB", + "jupyter": {}, + "notebookId": "676d00454d522334965beb55", + "runtime": { + "execution_status": null, + "is_visible": false, + "status": "default" + }, + "scrolled": false, + "slideshow": { + "slide_type": "slide" + }, + "tags": [] + }, "source": [ "模型2：部分池化模型（变化截距）" ] }, { "cell_type": "code", - "execution_count": 80, - "metadata": {}, - "outputs": [ - { - "name": "stderr", - "output_type": "stream", - "text": [ - "Auto-assigning NUTS sampler...\n", - "Initializing NUTS using jitter+adapt_diag...\n", - "Multiprocess sampling (4 chains in 4 jobs)\n", - "NUTS: [log_RTs_sigma, Intercept, Coherence, 1|subj_id_sigma, 1|subj_id_offset]\n" - ] + "execution_count": 15, + "metadata": { + "collapsed": false, + "id": "73B2A5D5CB7D412FB6C5564C77F0B771", + "jupyter": {}, + "notebookId": "676d00454d522334965beb55", + "scrolled": false, + "slideshow": { + "slide_type": "slide" }, + "tags": [], + "trusted": true + }, + "outputs": [ { "data": { - "application/vnd.jupyter.widget-view+json": { - "model_id": "6f25731d7fee487183a2cf9f9bbf3604", - "version_major": 2, - "version_minor": 0 - }, + "text/html": [ + "
Sampling 4 chains, 0 divergences ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╸━━  95% 0:00:02 / 0:00:33\n",
+       "
\n" + ], "text/plain": [ - "Output()" + "Sampling 4 chains, 0 divergences \u001b[38;2;23;100;244m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[38;2;23;100;244m╸\u001b[0m\u001b[38;5;237m━━\u001b[0m \u001b[35m 95%\u001b[0m \u001b[36m0:00:02\u001b[0m / \u001b[33m0:00:33\u001b[0m\n" ] }, "metadata": {}, @@ -738,8 +1027,7 @@ "name": "stderr", "output_type": "stream", "text": [ - "Sampling 4 chains for 1_000 tune and 1_000 draw iterations (4_000 + 4_000 draws total) took 44 seconds.\n", - "There were 196 divergences after tuning. Increase `target_accept` or reparameterize.\n", + "Sampling 4 chains for 1_000 tune and 728 draw iterations (4_000 + 2_912 draws total) took 33 seconds.\n", "The rhat statistic is larger than 1.01 for some parameters. This indicates problems during sampling. See https://arxiv.org/abs/1903.08008 for details\n", "The effective sample size per chain is smaller than 100 for some parameters. A higher number is needed for reliable rhat and ess computation. See https://arxiv.org/abs/1903.08008 for details\n" ] @@ -751,8 +1039,19 @@ }, { "cell_type": "code", - "execution_count": 50, - "metadata": {}, + "execution_count": 11, + "metadata": { + "collapsed": false, + "id": "7DC6FE9EB57444DB8B6182F604CA5A4F", + "jupyter": {}, + "notebookId": "676d00454d522334965beb55", + "scrolled": false, + "slideshow": { + "slide_type": "slide" + }, + "tags": [], + "trusted": true + }, "outputs": [ { "data": { @@ -788,142 +1087,492 @@ " \n", " \n", " \n", - " Intercept\n", - " 7.065\n", - " 0.309\n", - " 6.500\n", - " 7.688\n", - " 0.023\n", - " 0.016\n", - " 189.0\n", - " 272.0\n", - " 1.03\n", + " 1|subj_id[31727]\n", + " -0.096\n", + " 0.063\n", + " -0.210\n", + " 0.029\n", + " 0.006\n", + " 0.004\n", + " 121.0\n", + " 235.0\n", + " 1.04\n", " \n", " \n", - " Coherence\n", - " -0.063\n", - " 0.023\n", - " -0.108\n", - " -0.021\n", - " 0.000\n", - " 0.000\n", - " 2430.0\n", - " 2058.0\n", - " 1.01\n", + " 1|subj_id[71329]\n", + " -0.231\n", + " 0.063\n", + " -0.354\n", + " -0.115\n", + " 0.006\n", + " 0.004\n", + " 118.0\n", + " 230.0\n", + " 1.04\n", " \n", " \n", - " log_RTs_sigma\n", - " 0.559\n", - " 0.008\n", - " 0.544\n", - " 0.574\n", - " 0.000\n", - " 0.000\n", - " 1448.0\n", - " 2011.0\n", - " 1.00\n", + " 1|subj_id[71737]\n", + " 0.298\n", + " 0.064\n", + " 0.174\n", + " 0.417\n", + " 0.006\n", + " 0.004\n", + " 122.0\n", + " 249.0\n", + " 1.04\n", " \n", " \n", - " 1|subj_id_sigma\n", - " 0.700\n", - " 0.320\n", - " 0.251\n", - " 1.406\n", - " 0.059\n", - " 0.047\n", - " 49.0\n", - " 24.0\n", - " 1.06\n", + " 1|subj_id[75445]\n", + " 0.496\n", + " 0.065\n", + " 0.377\n", + " 0.623\n", + " 0.006\n", + " 0.004\n", + " 128.0\n", + " 243.0\n", + " 1.03\n", " \n", " \n", - " 1|subj_id[66670]\n", - " -0.139\n", - " 0.310\n", - " -0.735\n", - " 0.460\n", - " 0.023\n", - " 0.016\n", - " 193.0\n", - " 309.0\n", + " 1|subj_id[77704]\n", + " -0.204\n", + " 0.063\n", + " -0.318\n", + " -0.078\n", + " 0.006\n", + " 0.004\n", + " 125.0\n", + " 244.0\n", " 1.03\n", " \n", " \n", - " 1|subj_id[80941]\n", - " 0.137\n", - " 0.310\n", - " -0.477\n", - " 0.719\n", - " 0.023\n", - " 0.018\n", - " 188.0\n", - " 301.0\n", - " 1.03\n", + " 1|subj_id[79861]\n", + " -0.373\n", + " 0.062\n", + " -0.493\n", + " -0.256\n", + " 0.006\n", + " 0.004\n", + " 116.0\n", + " 181.0\n", + " 1.04\n", " \n", " \n", - " 1|subj_id[81844]\n", - " -0.641\n", - " 0.309\n", - " -1.269\n", - " -0.076\n", - " 0.023\n", - " 0.016\n", - " 189.0\n", - " 277.0\n", + " 1|subj_id[80035]\n", + " -0.159\n", + " 0.063\n", + " -0.274\n", + " -0.036\n", + " 0.006\n", + " 0.004\n", + " 121.0\n", + " 247.0\n", " 1.03\n", " \n", " \n", - " 1|subj_id[83824]\n", - " 0.085\n", - " 0.310\n", - " -0.502\n", - " 0.673\n", - " 0.023\n", - " 0.017\n", - " 186.0\n", - " 369.0\n", + " 1|subj_id[80362]\n", + " -0.123\n", + " 0.063\n", + " -0.240\n", + " -0.001\n", + " 0.006\n", + " 0.004\n", + " 122.0\n", + " 233.0\n", + " 1.04\n", + " \n", + " \n", + " 1|subj_id[80446]\n", + " -0.052\n", + " 0.063\n", + " -0.168\n", + " 0.073\n", + " 0.006\n", + " 0.004\n", + " 121.0\n", + " 233.0\n", " 1.03\n", " \n", " \n", - " 1|subj_id[83956]\n", - " 0.689\n", - " 0.311\n", - " 0.098\n", - " 1.282\n", - " 0.023\n", - " 0.018\n", - " 189.0\n", - " 312.0\n", + " 1|subj_id[80611]\n", + " 0.047\n", + " 0.063\n", + " -0.064\n", + " 0.177\n", + " 0.006\n", + " 0.004\n", + " 122.0\n", + " 244.0\n", " 1.03\n", " \n", - " \n", + " \n", + " 1|subj_id[80617]\n", + " 0.435\n", + " 0.064\n", + " 0.314\n", + " 0.561\n", + " 0.006\n", + " 0.004\n", + " 122.0\n", + " 266.0\n", + " 1.04\n", + " \n", + " \n", + " 1|subj_id[80620]\n", + " -0.207\n", + " 0.062\n", + " -0.320\n", + " -0.083\n", + " 0.006\n", + " 0.004\n", + " 118.0\n", + " 234.0\n", + " 1.04\n", + " \n", + " \n", + " 1|subj_id[80734]\n", + " -0.148\n", + " 0.063\n", + " -0.267\n", + " -0.028\n", + " 0.006\n", + " 0.004\n", + " 119.0\n", + " 216.0\n", + " 1.04\n", + " \n", + " \n", + " 1|subj_id[80977]\n", + " -0.075\n", + " 0.063\n", + " -0.196\n", + " 0.045\n", + " 0.006\n", + " 0.004\n", + " 122.0\n", + " 226.0\n", + " 1.04\n", + " \n", + " \n", + " 1|subj_id[81094]\n", + " -0.206\n", + " 0.063\n", + " -0.316\n", + " -0.076\n", + " 0.006\n", + " 0.004\n", + " 121.0\n", + " 227.0\n", + " 1.04\n", + " \n", + " \n", + " 1|subj_id[81211]\n", + " -0.060\n", + " 0.063\n", + " -0.179\n", + " 0.064\n", + " 0.006\n", + " 0.004\n", + " 120.0\n", + " 246.0\n", + " 1.04\n", + " \n", + " \n", + " 1|subj_id[81355]\n", + " -0.385\n", + " 0.063\n", + " -0.505\n", + " -0.262\n", + " 0.006\n", + " 0.004\n", + " 120.0\n", + " 202.0\n", + " 1.04\n", + " \n", + " \n", + " 1|subj_id[81376]\n", + " -0.309\n", + " 0.062\n", + " -0.419\n", + " -0.184\n", + " 0.006\n", + " 0.004\n", + " 121.0\n", + " 230.0\n", + " 1.04\n", + " \n", + " \n", + " 1|subj_id[81478]\n", + " -0.325\n", + " 0.062\n", + " -0.440\n", + " -0.201\n", + " 0.006\n", + " 0.004\n", + " 117.0\n", + " 216.0\n", + " 1.04\n", + " \n", + " \n", + " 1|subj_id[81487]\n", + " 0.098\n", + " 0.064\n", + " -0.028\n", + " 0.217\n", + " 0.006\n", + " 0.004\n", + " 120.0\n", + " 244.0\n", + " 1.04\n", + " \n", + " \n", + " 1|subj_id[81577]\n", + " -0.049\n", + " 0.064\n", + " -0.177\n", + " 0.068\n", + " 0.006\n", + " 0.004\n", + " 118.0\n", + " 214.0\n", + " 1.04\n", + " \n", + " \n", + " 1|subj_id[81586]\n", + " 0.145\n", + " 0.064\n", + " 0.031\n", + " 0.276\n", + " 0.006\n", + " 0.004\n", + " 125.0\n", + " 240.0\n", + " 1.03\n", + " \n", + " \n", + " 1|subj_id[81709]\n", + " -0.363\n", + " 0.063\n", + " -0.477\n", + " -0.238\n", + " 0.006\n", + " 0.004\n", + " 119.0\n", + " 201.0\n", + " 1.04\n", + " \n", + " \n", + " 1|subj_id[82111]\n", + " -0.031\n", + " 0.064\n", + " -0.154\n", + " 0.088\n", + " 0.006\n", + " 0.004\n", + " 119.0\n", + " 251.0\n", + " 1.04\n", + " \n", + " \n", + " 1|subj_id[83824]\n", + " 0.210\n", + " 0.065\n", + " 0.091\n", + " 0.339\n", + " 0.006\n", + " 0.004\n", + " 130.0\n", + " 273.0\n", + " 1.03\n", + " \n", + " \n", + " 1|subj_id[83953]\n", + " -0.016\n", + " 0.064\n", + " -0.138\n", + " 0.104\n", + " 0.006\n", + " 0.004\n", + " 121.0\n", + " 206.0\n", + " 1.04\n", + " \n", + " \n", + " 1|subj_id[83956]\n", + " 0.810\n", + " 0.066\n", + " 0.692\n", + " 0.941\n", + " 0.006\n", + " 0.004\n", + " 127.0\n", + " 336.0\n", + " 1.03\n", + " \n", + " \n", + " 1|subj_id[84304]\n", + " 0.440\n", + " 0.064\n", + " 0.318\n", + " 0.564\n", + " 0.006\n", + " 0.004\n", + " 125.0\n", + " 226.0\n", + " 1.04\n", + " \n", + " \n", + " 1|subj_id[84322]\n", + " 0.091\n", + " 0.062\n", + " -0.022\n", + " 0.214\n", + " 0.006\n", + " 0.004\n", + " 113.0\n", + " 212.0\n", + " 1.04\n", + " \n", + " \n", + " 1|subj_id[84370]\n", + " 0.610\n", + " 0.066\n", + " 0.492\n", + " 0.740\n", + " 0.006\n", + " 0.004\n", + " 133.0\n", + " 274.0\n", + " 1.03\n", + " \n", + " \n", + " 1|subj_id_sigma\n", + " 0.321\n", + " 0.045\n", + " 0.242\n", + " 0.402\n", + " 0.003\n", + " 0.002\n", + " 204.0\n", + " 398.0\n", + " 1.01\n", + " \n", + " \n", + " Coherence\n", + " -0.101\n", + " 0.008\n", + " -0.117\n", + " -0.087\n", + " 0.000\n", + " 0.000\n", + " 1452.0\n", + " 2032.0\n", + " 1.00\n", + " \n", + " \n", + " Intercept\n", + " 6.960\n", + " 0.060\n", + " 6.840\n", + " 7.068\n", + " 0.006\n", + " 0.004\n", + " 108.0\n", + " 194.0\n", + " 1.04\n", + " \n", + " \n", + " log_RTs_sigma\n", + " 0.495\n", + " 0.003\n", + " 0.490\n", + " 0.500\n", + " 0.000\n", + " 0.000\n", + " 1540.0\n", + " 2239.0\n", + " 1.00\n", + " \n", + " \n", "\n", "" ], "text/plain": [ " mean sd hdi_3% hdi_97% mcse_mean mcse_sd ess_bulk \\\n", - "Intercept 7.065 0.309 6.500 7.688 0.023 0.016 189.0 \n", - "Coherence -0.063 0.023 -0.108 -0.021 0.000 0.000 2430.0 \n", - "log_RTs_sigma 0.559 0.008 0.544 0.574 0.000 0.000 1448.0 \n", - "1|subj_id_sigma 0.700 0.320 0.251 1.406 0.059 0.047 49.0 \n", - "1|subj_id[66670] -0.139 0.310 -0.735 0.460 0.023 0.016 193.0 \n", - "1|subj_id[80941] 0.137 0.310 -0.477 0.719 0.023 0.018 188.0 \n", - "1|subj_id[81844] -0.641 0.309 -1.269 -0.076 0.023 0.016 189.0 \n", - "1|subj_id[83824] 0.085 0.310 -0.502 0.673 0.023 0.017 186.0 \n", - "1|subj_id[83956] 0.689 0.311 0.098 1.282 0.023 0.018 189.0 \n", + "1|subj_id[31727] -0.096 0.063 -0.210 0.029 0.006 0.004 121.0 \n", + "1|subj_id[71329] -0.231 0.063 -0.354 -0.115 0.006 0.004 118.0 \n", + "1|subj_id[71737] 0.298 0.064 0.174 0.417 0.006 0.004 122.0 \n", + "1|subj_id[75445] 0.496 0.065 0.377 0.623 0.006 0.004 128.0 \n", + "1|subj_id[77704] -0.204 0.063 -0.318 -0.078 0.006 0.004 125.0 \n", + "1|subj_id[79861] -0.373 0.062 -0.493 -0.256 0.006 0.004 116.0 \n", + "1|subj_id[80035] -0.159 0.063 -0.274 -0.036 0.006 0.004 121.0 \n", + "1|subj_id[80362] -0.123 0.063 -0.240 -0.001 0.006 0.004 122.0 \n", + "1|subj_id[80446] -0.052 0.063 -0.168 0.073 0.006 0.004 121.0 \n", + "1|subj_id[80611] 0.047 0.063 -0.064 0.177 0.006 0.004 122.0 \n", + "1|subj_id[80617] 0.435 0.064 0.314 0.561 0.006 0.004 122.0 \n", + "1|subj_id[80620] -0.207 0.062 -0.320 -0.083 0.006 0.004 118.0 \n", + "1|subj_id[80734] -0.148 0.063 -0.267 -0.028 0.006 0.004 119.0 \n", + "1|subj_id[80977] -0.075 0.063 -0.196 0.045 0.006 0.004 122.0 \n", + "1|subj_id[81094] -0.206 0.063 -0.316 -0.076 0.006 0.004 121.0 \n", + "1|subj_id[81211] -0.060 0.063 -0.179 0.064 0.006 0.004 120.0 \n", + "1|subj_id[81355] -0.385 0.063 -0.505 -0.262 0.006 0.004 120.0 \n", + "1|subj_id[81376] -0.309 0.062 -0.419 -0.184 0.006 0.004 121.0 \n", + "1|subj_id[81478] -0.325 0.062 -0.440 -0.201 0.006 0.004 117.0 \n", + "1|subj_id[81487] 0.098 0.064 -0.028 0.217 0.006 0.004 120.0 \n", + "1|subj_id[81577] -0.049 0.064 -0.177 0.068 0.006 0.004 118.0 \n", + "1|subj_id[81586] 0.145 0.064 0.031 0.276 0.006 0.004 125.0 \n", + "1|subj_id[81709] -0.363 0.063 -0.477 -0.238 0.006 0.004 119.0 \n", + "1|subj_id[82111] -0.031 0.064 -0.154 0.088 0.006 0.004 119.0 \n", + "1|subj_id[83824] 0.210 0.065 0.091 0.339 0.006 0.004 130.0 \n", + "1|subj_id[83953] -0.016 0.064 -0.138 0.104 0.006 0.004 121.0 \n", + "1|subj_id[83956] 0.810 0.066 0.692 0.941 0.006 0.004 127.0 \n", + "1|subj_id[84304] 0.440 0.064 0.318 0.564 0.006 0.004 125.0 \n", + "1|subj_id[84322] 0.091 0.062 -0.022 0.214 0.006 0.004 113.0 \n", + "1|subj_id[84370] 0.610 0.066 0.492 0.740 0.006 0.004 133.0 \n", + "1|subj_id_sigma 0.321 0.045 0.242 0.402 0.003 0.002 204.0 \n", + "Coherence -0.101 0.008 -0.117 -0.087 0.000 0.000 1452.0 \n", + "Intercept 6.960 0.060 6.840 7.068 0.006 0.004 108.0 \n", + "log_RTs_sigma 0.495 0.003 0.490 0.500 0.000 0.000 1540.0 \n", "\n", " ess_tail r_hat \n", - "Intercept 272.0 1.03 \n", - "Coherence 2058.0 1.01 \n", - "log_RTs_sigma 2011.0 1.00 \n", - "1|subj_id_sigma 24.0 1.06 \n", - "1|subj_id[66670] 309.0 1.03 \n", - "1|subj_id[80941] 301.0 1.03 \n", - "1|subj_id[81844] 277.0 1.03 \n", - "1|subj_id[83824] 369.0 1.03 \n", - "1|subj_id[83956] 312.0 1.03 " + "1|subj_id[31727] 235.0 1.04 \n", + "1|subj_id[71329] 230.0 1.04 \n", + "1|subj_id[71737] 249.0 1.04 \n", + "1|subj_id[75445] 243.0 1.03 \n", + "1|subj_id[77704] 244.0 1.03 \n", + "1|subj_id[79861] 181.0 1.04 \n", + "1|subj_id[80035] 247.0 1.03 \n", + "1|subj_id[80362] 233.0 1.04 \n", + "1|subj_id[80446] 233.0 1.03 \n", + "1|subj_id[80611] 244.0 1.03 \n", + "1|subj_id[80617] 266.0 1.04 \n", + "1|subj_id[80620] 234.0 1.04 \n", + "1|subj_id[80734] 216.0 1.04 \n", + "1|subj_id[80977] 226.0 1.04 \n", + "1|subj_id[81094] 227.0 1.04 \n", + "1|subj_id[81211] 246.0 1.04 \n", + "1|subj_id[81355] 202.0 1.04 \n", + "1|subj_id[81376] 230.0 1.04 \n", + "1|subj_id[81478] 216.0 1.04 \n", + "1|subj_id[81487] 244.0 1.04 \n", + "1|subj_id[81577] 214.0 1.04 \n", + "1|subj_id[81586] 240.0 1.03 \n", + "1|subj_id[81709] 201.0 1.04 \n", + "1|subj_id[82111] 251.0 1.04 \n", + "1|subj_id[83824] 273.0 1.03 \n", + "1|subj_id[83953] 206.0 1.04 \n", + "1|subj_id[83956] 336.0 1.03 \n", + "1|subj_id[84304] 226.0 1.04 \n", + "1|subj_id[84322] 212.0 1.04 \n", + "1|subj_id[84370] 274.0 1.03 \n", + "1|subj_id_sigma 398.0 1.01 \n", + "Coherence 2032.0 1.00 \n", + "Intercept 194.0 1.04 \n", + "log_RTs_sigma 2239.0 1.00 " ] }, - "execution_count": 50, + "execution_count": 11, "metadata": {}, "output_type": "execute_result" } @@ -935,24 +1584,37 @@ }, { "cell_type": "code", - "execution_count": 51, - "metadata": {}, + "execution_count": 12, + "metadata": { + "collapsed": false, + "id": "7638987AAE8B4B5D8FAF66D11DA938C2", + "jupyter": {}, + "notebookId": "676d00454d522334965beb55", + "scrolled": false, + "slideshow": { + "slide_type": "slide" + }, + "tags": [], + "trusted": true + }, "outputs": [ { "data": { "text/plain": [ - "array([], dtype=object)" + "array([], dtype=object)" ] }, - "execution_count": 51, + "execution_count": 12, "metadata": {}, "output_type": "execute_result" }, { "data": { - "image/png": "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", + "text/html": [ + "" + ], "text/plain": [ - "
" + "
" ] }, "metadata": {}, @@ -968,35 +1630,49 @@ }, { "cell_type": "markdown", - "metadata": {}, + "metadata": { + "id": "96C3F819D5CA42958EFAF19536FFE066", + "jupyter": {}, + "notebookId": "676d00454d522334965beb55", + "runtime": { + "execution_status": null, + "is_visible": false, + "status": "default" + }, + "scrolled": false, + "slideshow": { + "slide_type": "slide" + }, + "tags": [] + }, "source": [ "**模型3**：部分池化模型（变化截距和斜率）" ] }, { "cell_type": "code", - "execution_count": 81, - "metadata": {}, - "outputs": [ - { - "name": "stderr", - "output_type": "stream", - "text": [ - "Auto-assigning NUTS sampler...\n", - "Initializing NUTS using jitter+adapt_diag...\n", - "Multiprocess sampling (4 chains in 4 jobs)\n", - "NUTS: [log_RTs_sigma, Intercept, Coherence, 1|subj_id_sigma, 1|subj_id_offset, Coherence|subj_id_sigma, Coherence|subj_id_offset]\n" - ] + "execution_count": 5, + "metadata": { + "collapsed": false, + "id": "12B4A957ED8F4B6A9DDE4FE0F782D9BB", + "jupyter": {}, + "notebookId": "676d00454d522334965beb55", + "scrolled": false, + "slideshow": { + "slide_type": "slide" }, + "tags": [], + "trusted": true + }, + "outputs": [ { "data": { - "application/vnd.jupyter.widget-view+json": { - "model_id": "a625715452854b0097955e21cac11498", - "version_major": 2, - "version_minor": 0 - }, + "text/html": [ + "
Sampling 4 chains, 0 divergences ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 100% 0:00:00 / 0:01:54\n",
+       "
\n" + ], "text/plain": [ - "Output()" + "Sampling 4 chains, 0 divergences \u001b[32m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[35m100%\u001b[0m \u001b[36m0:00:00\u001b[0m / \u001b[33m0:01:54\u001b[0m\n" ] }, "metadata": {}, @@ -1029,10 +1705,7 @@ "name": "stderr", "output_type": "stream", "text": [ - "Sampling 4 chains for 1_000 tune and 1_000 draw iterations (4_000 + 4_000 draws total) took 76 seconds.\n", - "There were 308 divergences after tuning. Increase `target_accept` or reparameterize.\n", - "The rhat statistic is larger than 1.01 for some parameters. This indicates problems during sampling. See https://arxiv.org/abs/1903.08008 for details\n", - "The effective sample size per chain is smaller than 100 for some parameters. A higher number is needed for reliable rhat and ess computation. See https://arxiv.org/abs/1903.08008 for details\n" + "Sampling 4 chains for 1_000 tune and 1_000 draw iterations (4_000 + 4_000 draws total) took 115 seconds.\n" ] } ], @@ -1042,8 +1715,19 @@ }, { "cell_type": "code", - "execution_count": 53, - "metadata": {}, + "execution_count": 6, + "metadata": { + "collapsed": false, + "id": "E528E6418FBE406E8D6042B6A2569AD4", + "jupyter": {}, + "notebookId": "676d00454d522334965beb55", + "scrolled": false, + "slideshow": { + "slide_type": "slide" + }, + "tags": [], + "trusted": true + }, "outputs": [ { "data": { @@ -1076,2896 +1760,331 @@ " ess_tail\n", " r_hat\n", " \n", - " \n", - " \n", - " \n", - " Intercept\n", - " 7.080\n", - " 0.332\n", - " 6.512\n", - " 7.798\n", - " 0.012\n", - " 0.009\n", - " 716.0\n", - " 849.0\n", - " 1.00\n", - " \n", - " \n", - " Coherence\n", - " -0.097\n", - " 0.124\n", - " -0.308\n", - " 0.119\n", - " 0.005\n", - " 0.003\n", - " 966.0\n", - " 1070.0\n", - " 1.01\n", - " \n", - " \n", - " log_RTs_sigma\n", - " 0.556\n", - " 0.008\n", - " 0.541\n", - " 0.571\n", - " 0.000\n", - " 0.000\n", - " 1333.0\n", - " 1339.0\n", - " 1.00\n", - " \n", - " \n", - " 1|subj_id_sigma\n", - " 0.785\n", - " 0.377\n", - " 0.313\n", - " 1.555\n", - " 0.028\n", - " 0.024\n", - " 323.0\n", - " 146.0\n", - " 1.01\n", - " \n", - " \n", - " Coherence|subj_id_sigma\n", - " 0.215\n", - " 0.151\n", - " 0.048\n", - " 0.449\n", - " 0.006\n", - " 0.005\n", - " 666.0\n", - " 980.0\n", - " 1.01\n", - " \n", - " \n", - " 1|subj_id[66670]\n", - " -0.201\n", - " 0.334\n", - " -0.888\n", - " 0.395\n", - " 0.013\n", - " 0.009\n", - " 713.0\n", - " 855.0\n", - " 1.00\n", - " \n", - " \n", - " 1|subj_id[80941]\n", - " 0.168\n", - " 0.333\n", - " -0.489\n", - " 0.785\n", - " 0.013\n", - " 0.009\n", - " 714.0\n", - " 928.0\n", - " 1.00\n", - " \n", - " \n", - " 1|subj_id[81844]\n", - " -0.692\n", - " 0.333\n", - " -1.404\n", - " -0.115\n", - " 0.013\n", - " 0.009\n", - " 718.0\n", - " 899.0\n", - " 1.00\n", - " \n", - " \n", - " 1|subj_id[83824]\n", - " 0.070\n", - " 0.333\n", - " -0.610\n", - " 0.670\n", - " 0.013\n", - " 0.009\n", - " 703.0\n", - " 715.0\n", - " 1.00\n", - " \n", - " \n", - " 1|subj_id[83956]\n", - " 0.786\n", - " 0.334\n", - " 0.114\n", - " 1.388\n", - " 0.012\n", - " 0.009\n", - " 731.0\n", - " 998.0\n", - " 1.00\n", - " \n", - " \n", - " Coherence|subj_id[66670]\n", - " 0.122\n", - " 0.130\n", - " -0.114\n", - " 0.327\n", - " 0.005\n", - " 0.003\n", - " 1068.0\n", - " 1224.0\n", - " 1.00\n", - " \n", - " \n", - " Coherence|subj_id[80941]\n", - " -0.061\n", - " 0.131\n", - " -0.289\n", - " 0.161\n", - " 0.005\n", - " 0.004\n", - " 1126.0\n", - " 1060.0\n", - " 1.01\n", - " \n", - " \n", - " Coherence|subj_id[81844]\n", - " 0.107\n", - " 0.127\n", - " -0.126\n", - " 0.316\n", - " 0.005\n", - " 0.003\n", - " 998.0\n", - " 1155.0\n", - " 1.00\n", - " \n", - " \n", - " Coherence|subj_id[83824]\n", - " 0.034\n", - " 0.131\n", - " -0.191\n", - " 0.257\n", - " 0.005\n", - " 0.004\n", - " 1125.0\n", - " 1243.0\n", - " 1.00\n", - " \n", - " \n", - " Coherence|subj_id[83956]\n", - " -0.182\n", - " 0.132\n", - " -0.405\n", - " 0.043\n", - " 0.005\n", - " 0.004\n", - " 1070.0\n", - " 1167.0\n", - " 1.00\n", - " \n", - " \n", - "\n", - "" - ], - "text/plain": [ - " mean sd hdi_3% hdi_97% mcse_mean mcse_sd \\\n", - "Intercept 7.080 0.332 6.512 7.798 0.012 0.009 \n", - "Coherence -0.097 0.124 -0.308 0.119 0.005 0.003 \n", - "log_RTs_sigma 0.556 0.008 0.541 0.571 0.000 0.000 \n", - "1|subj_id_sigma 0.785 0.377 0.313 1.555 0.028 0.024 \n", - "Coherence|subj_id_sigma 0.215 0.151 0.048 0.449 0.006 0.005 \n", - "1|subj_id[66670] -0.201 0.334 -0.888 0.395 0.013 0.009 \n", - "1|subj_id[80941] 0.168 0.333 -0.489 0.785 0.013 0.009 \n", - "1|subj_id[81844] -0.692 0.333 -1.404 -0.115 0.013 0.009 \n", - "1|subj_id[83824] 0.070 0.333 -0.610 0.670 0.013 0.009 \n", - "1|subj_id[83956] 0.786 0.334 0.114 1.388 0.012 0.009 \n", - "Coherence|subj_id[66670] 0.122 0.130 -0.114 0.327 0.005 0.003 \n", - "Coherence|subj_id[80941] -0.061 0.131 -0.289 0.161 0.005 0.004 \n", - "Coherence|subj_id[81844] 0.107 0.127 -0.126 0.316 0.005 0.003 \n", - "Coherence|subj_id[83824] 0.034 0.131 -0.191 0.257 0.005 0.004 \n", - "Coherence|subj_id[83956] -0.182 0.132 -0.405 0.043 0.005 0.004 \n", - "\n", - " ess_bulk ess_tail r_hat \n", - "Intercept 716.0 849.0 1.00 \n", - "Coherence 966.0 1070.0 1.01 \n", - "log_RTs_sigma 1333.0 1339.0 1.00 \n", - "1|subj_id_sigma 323.0 146.0 1.01 \n", - "Coherence|subj_id_sigma 666.0 980.0 1.01 \n", - "1|subj_id[66670] 713.0 855.0 1.00 \n", - "1|subj_id[80941] 714.0 928.0 1.00 \n", - "1|subj_id[81844] 718.0 899.0 1.00 \n", - "1|subj_id[83824] 703.0 715.0 1.00 \n", - "1|subj_id[83956] 731.0 998.0 1.00 \n", - "Coherence|subj_id[66670] 1068.0 1224.0 1.00 \n", - "Coherence|subj_id[80941] 1126.0 1060.0 1.01 \n", - "Coherence|subj_id[81844] 998.0 1155.0 1.00 \n", - "Coherence|subj_id[83824] 1125.0 1243.0 1.00 \n", - "Coherence|subj_id[83956] 1070.0 1167.0 1.00 " - ] - }, - "execution_count": 53, - "metadata": {}, - "output_type": "execute_result" - } - ], - "source": [ - "var_both_para = az.summary(var_both_trace)\n", - "var_both_para" - ] - }, - { - "cell_type": "code", - "execution_count": 63, - "metadata": {}, - "outputs": [ - { - "data": { - "image/png": "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", - "text/plain": [ - "
" - ] - }, - "metadata": {}, - "output_type": "display_data" - } - ], - "source": [ - "# 设置绘图坐标\n", - "figs, (ax1, ax2) = plt.subplots(1,2, figsize = (20,5))\n", - "# 绘制变化的截距\n", - "az.plot_forest(var_both_trace,\n", - " var_names=[\"Coherence\\\\|subj_id\", \"~sigma\", \"~1|\", \"~Intercept\"],\n", - " filter_vars=\"like\",\n", - " combined = True,\n", - " ax=ax1)\n", - "# 绘制变化的斜率\n", - "az.plot_forest(var_both_trace,\n", - " var_names=[\"1|subj_id\", \"~sigma\", \"~Coherence\", \"~Intercept\"],\n", - " filter_vars=\"like\",\n", - " combined = True,\n", - " ax=ax2)\n", - "plt.show()" - ] - }, - { - "cell_type": "markdown", - "metadata": {}, - "source": [ - "### 模型比较\n", - "\n", - "##### 模型评估指标\n", - "\n", - "在分析模型的预测能力时，有绝对指标和相对指标，绝对指标用于衡量模型预测值与真实值之间的差异，相对指标用于比较不同模型的预测能力，通常用于不同方法或模型之间的性能对比。\n", - "\n", - "##### 绝对指标：\n", - "\n", - "* 在之前的课程中介绍过对后验预测结果进行评估的两种方法 \n", - "\n", - "* 一是**MAE**，即后验预测值与真实值之间预测误差的中位数，二是**within_95**，即真实值是否落在95%后验预测区间内 \n", - "\n", - "* 在这里调用之前写过的计算两种指标的方法，评估两种分层模型的后验预测结果\n", - "\n", - "\n", - "##### 相对指标\n", - "\n", - "在实际操作中，我们通过 `ArViz` 的函数`az.loo`计算 $ELPD_{LOO-CV}$。 \n", - "\n", - "PSIS-LOO-CV 有两大优势： \n", - "1. 计算速度快，且结果稳健 \n", - "2. 提供了丰富的模型诊断指标 " - ] - }, - { - "cell_type": "code", - "execution_count": 90, - "metadata": {}, - "outputs": [], - "source": [ - "complete_pooled_model.predict(complete_pooled_trace, kind=\"pps\")\n", - "var_inter_model.predict(var_inter_trace, kind=\"pps\")\n", - "var_both_model.predict(var_both_trace, kind=\"pps\")" - ] - }, - { - "cell_type": "code", - "execution_count": 92, - "metadata": {}, - "outputs": [ - { - "data": { - "text/html": [ - "\n", - "
\n", - "
\n", - "
arviz.InferenceData
\n", - "
\n", - "
    \n", - " \n", - "
  • \n", - " \n", - " \n", - "
    \n", - "
    \n", - "
      \n", - "
      \n", - "\n", - "\n", - "\n", - "\n", - "\n", - "\n", - "\n", - "\n", - "\n", - "\n", - "\n", - "\n", - "\n", - "\n", - "
      <xarray.Dataset>\n",
-       "Dimensions:        (chain: 4, draw: 1000, log_RTs_obs: 2326)\n",
-       "Coordinates:\n",
-       "  * chain          (chain) int32 0 1 2 3\n",
-       "  * draw           (draw) int32 0 1 2 3 4 5 6 7 ... 993 994 995 996 997 998 999\n",
-       "  * log_RTs_obs    (log_RTs_obs) int32 0 1 2 3 4 5 ... 2321 2322 2323 2324 2325\n",
-       "Data variables:\n",
-       "    Intercept      (chain, draw) float64 6.948 6.952 6.953 ... 6.977 6.911 6.941\n",
-       "    Coherence      (chain, draw) float64 -0.05112 -0.09796 ... -0.03254 -0.05348\n",
-       "    log_RTs_sigma  (chain, draw) float64 0.6968 0.7148 0.7104 ... 0.6805 0.6917\n",
-       "    log_RTs_mean   (chain, draw, log_RTs_obs) float64 6.948 6.948 ... 6.887\n",
-       "Attributes:\n",
-       "    created_at:                  2024-12-26T03:18:17.443034\n",
-       "    arviz_version:               0.17.1\n",
-       "    inference_library:           pymc\n",
-       "    inference_library_version:   5.16.2\n",
-       "    sampling_time:               19.907405138015747\n",
-       "    tuning_steps:                1000\n",
-       "    modeling_interface:          bambi\n",
-       "    modeling_interface_version:  0.13.0

      \n", - "
    \n", - "
    \n", - "
    • \n", - " \n", - "
  • \n", - " \n", - " \n", - "
    \n", - "
    \n", - "
      \n", - "
      \n", - "\n", - "\n", - "\n", - "\n", - "\n", - "\n", - "\n", - "\n", - "\n", - "\n", - "\n", - "\n", - "\n", - "\n", - "
      <xarray.Dataset>\n",
-       "Dimensions:      (chain: 4, draw: 1000, log_RTs_obs: 2326)\n",
-       "Coordinates:\n",
-       "  * chain        (chain) int32 0 1 2 3\n",
-       "  * draw         (draw) int32 0 1 2 3 4 5 6 7 ... 993 994 995 996 997 998 999\n",
-       "  * log_RTs_obs  (log_RTs_obs) int32 0 1 2 3 4 5 ... 2321 2322 2323 2324 2325\n",
-       "Data variables:\n",
-       "    log_RTs      (chain, draw, log_RTs_obs) float64 6.407 6.604 ... 7.459 8.231\n",
-       "Attributes:\n",
-       "    modeling_interface:          bambi\n",
-       "    modeling_interface_version:  0.13.0

      \n", - "
    \n", - "
    \n", - "
    • \n", - " \n", - "
  • \n", - " \n", - " \n", - "
    \n", - "
    \n", - "
      \n", - "
      \n", - "\n", - "\n", - "\n", - "\n", - "\n", - "\n", - "\n", - "\n", - "\n", - "\n", - "\n", - "\n", - "\n", - "\n", - "
      <xarray.Dataset>\n",
-       "Dimensions:      (chain: 4, draw: 1000, log_RTs_obs: 2326)\n",
-       "Coordinates:\n",
-       "  * chain        (chain) int32 0 1 2 3\n",
-       "  * draw         (draw) int32 0 1 2 3 4 5 6 7 ... 993 994 995 996 997 998 999\n",
-       "  * log_RTs_obs  (log_RTs_obs) int32 0 1 2 3 4 5 ... 2321 2322 2323 2324 2325\n",
-       "Data variables:\n",
-       "    log_RTs      (chain, draw, log_RTs_obs) float64 -0.6775 -0.762 ... -0.9516\n",
-       "Attributes:\n",
-       "    created_at:                  2024-12-26T03:18:17.693780\n",
-       "    arviz_version:               0.17.1\n",
-       "    inference_library:           pymc\n",
-       "    inference_library_version:   5.16.2\n",
-       "    modeling_interface:          bambi\n",
-       "    modeling_interface_version:  0.13.0

      \n", - "
    \n", - "
    \n", - "
    • \n", - " \n", - "
  • \n", - " \n", - " \n", - "
    \n", - "
    \n", - "
      \n", - "
      \n", - "\n", - "\n", - "\n", - "\n", - "\n", - "\n", - "\n", - "\n", - "\n", - "\n", - "\n", - "\n", - "\n", - "\n", - "
      <xarray.Dataset>\n",
-       "Dimensions:                (chain: 4, draw: 1000)\n",
-       "Coordinates:\n",
-       "  * chain                  (chain) int32 0 1 2 3\n",
-       "  * draw                   (draw) int32 0 1 2 3 4 5 ... 994 995 996 997 998 999\n",
-       "Data variables: (12/17)\n",
-       "    perf_counter_start     (chain, draw) float64 9.687e+04 ... 9.688e+04\n",
-       "    step_size              (chain, draw) float64 0.8867 0.8867 ... 0.9589 0.9589\n",
-       "    diverging              (chain, draw) bool False False False ... False False\n",
-       "    lp                     (chain, draw) float64 -2.505e+03 ... -2.505e+03\n",
-       "    smallest_eigval        (chain, draw) float64 nan nan nan nan ... nan nan nan\n",
-       "    energy_error           (chain, draw) float64 0.0 0.3789 ... 1.379 -1.532\n",
-       "    ...                     ...\n",
-       "    tree_depth             (chain, draw) int64 1 2 1 2 2 2 2 2 ... 1 2 2 2 2 2 1\n",
-       "    energy                 (chain, draw) float64 2.506e+03 ... 2.509e+03\n",
-       "    max_energy_error       (chain, draw) float64 0.817 0.7752 ... 1.835 -1.532\n",
-       "    reached_max_treedepth  (chain, draw) bool False False False ... False False\n",
-       "    process_time_diff      (chain, draw) float64 0.0 0.0 0.0 0.0 ... 0.0 0.0 0.0\n",
-       "    acceptance_rate        (chain, draw) float64 0.4418 0.6464 ... 0.2306 1.0\n",
-       "Attributes:\n",
-       "    created_at:                  2024-12-26T03:18:17.455046\n",
-       "    arviz_version:               0.17.1\n",
-       "    inference_library:           pymc\n",
-       "    inference_library_version:   5.16.2\n",
-       "    sampling_time:               19.907405138015747\n",
-       "    tuning_steps:                1000\n",
-       "    modeling_interface:          bambi\n",
-       "    modeling_interface_version:  0.13.0

      \n", - "
    \n", - "
    \n", - "
    • \n", - " \n", - "
  • \n", - " \n", - " \n", - "
    \n", - "
    \n", - "
      \n", - "
      \n", - "\n", - "\n", - "\n", - "\n", - "\n", - "\n", - "\n", - "\n", - "\n", - "\n", - "\n", - "\n", - "\n", - "\n", - "
      <xarray.Dataset>\n",
-       "Dimensions:      (log_RTs_obs: 2326)\n",
-       "Coordinates:\n",
-       "  * log_RTs_obs  (log_RTs_obs) int32 0 1 2 3 4 5 ... 2321 2322 2323 2324 2325\n",
-       "Data variables:\n",
-       "    log_RTs      (log_RTs_obs) float64 7.29 7.394 7.788 ... 8.1 7.757 7.507\n",
-       "Attributes:\n",
-       "    created_at:                  2024-12-26T03:18:17.461051\n",
-       "    arviz_version:               0.17.1\n",
-       "    inference_library:           pymc\n",
-       "    inference_library_version:   5.16.2\n",
-       "    modeling_interface:          bambi\n",
-       "    modeling_interface_version:  0.13.0

      \n", - "
    \n", - "
    \n", - "
    • \n", - " \n", - "
\n", - "
\n", - " " + " \n", + " \n", + " \n", + " 1|subj_id[31727]\n", + " -0.116\n", + " 0.069\n", + " -0.244\n", + " 0.014\n", + " 0.003\n", + " 0.002\n", + " 611.0\n", + " 974.0\n", + " 1.01\n", + " \n", + " \n", + " 1|subj_id[71329]\n", + " -0.270\n", + " 0.070\n", + " -0.408\n", + " -0.148\n", + " 0.003\n", + " 0.002\n", + " 574.0\n", + " 917.0\n", + " 1.01\n", + " \n", + " \n", + " 1|subj_id[71737]\n", + " 0.334\n", + " 0.070\n", + " 0.205\n", + " 0.467\n", + " 0.003\n", + " 0.002\n", + " 588.0\n", + " 1095.0\n", + " 1.01\n", + " \n", + " \n", + " 1|subj_id[75445]\n", + " 0.439\n", + " 0.072\n", + " 0.310\n", + " 0.580\n", + " 0.003\n", + " 0.002\n", + " 633.0\n", + " 1069.0\n", + " 1.01\n", + " \n", + " \n", + " 1|subj_id[77704]\n", + " -0.270\n", + " 0.070\n", + " -0.399\n", + " -0.138\n", + " 0.003\n", + " 0.002\n", + " 581.0\n", + " 1041.0\n", + " 1.01\n", + " \n", + " \n", + " ...\n", + " ...\n", + " ...\n", + " ...\n", + " ...\n", + " ...\n", + " ...\n", + " ...\n", + " ...\n", + " ...\n", + " \n", + " \n", + " Coherence|subj_id[84322]\n", + " 0.044\n", + " 0.033\n", + " -0.017\n", + " 0.105\n", + " 0.001\n", + " 0.000\n", + " 2958.0\n", + " 2790.0\n", + " 1.00\n", + " \n", + " \n", + " Coherence|subj_id[84370]\n", + " -0.129\n", + " 0.051\n", + " -0.222\n", + " -0.031\n", + " 0.001\n", + " 0.001\n", + " 4995.0\n", + " 2613.0\n", + " 1.00\n", + " \n", + " \n", + " Coherence|subj_id_sigma\n", + " 0.090\n", + " 0.016\n", + " 0.063\n", + " 0.122\n", + " 0.000\n", + " 0.000\n", + " 1571.0\n", + " 2345.0\n", + " 1.00\n", + " \n", + " \n", + " Intercept\n", + " 6.977\n", + " 0.064\n", + " 6.857\n", + " 7.093\n", + " 0.003\n", + " 0.002\n", + " 508.0\n", + " 737.0\n", + " 1.01\n", + " \n", + " \n", + " log_RTs_sigma\n", + " 0.493\n", + " 0.003\n", + " 0.488\n", + " 0.498\n", + " 0.000\n", + " 0.000\n", + " 6013.0\n", + " 2276.0\n", + " 1.00\n", + " \n", + " \n", + "\n", + "

65 rows × 9 columns

\n", + "" ], "text/plain": [ - "Inference data with groups:\n", - "\t> posterior\n", - "\t> posterior_predictive\n", - "\t> log_likelihood\n", - "\t> sample_stats\n", - "\t> observed_data" + " mean sd hdi_3% hdi_97% mcse_mean mcse_sd \\\n", + "1|subj_id[31727] -0.116 0.069 -0.244 0.014 0.003 0.002 \n", + "1|subj_id[71329] -0.270 0.070 -0.408 -0.148 0.003 0.002 \n", + "1|subj_id[71737] 0.334 0.070 0.205 0.467 0.003 0.002 \n", + "1|subj_id[75445] 0.439 0.072 0.310 0.580 0.003 0.002 \n", + "1|subj_id[77704] -0.270 0.070 -0.399 -0.138 0.003 0.002 \n", + "... ... ... ... ... ... ... \n", + "Coherence|subj_id[84322] 0.044 0.033 -0.017 0.105 0.001 0.000 \n", + "Coherence|subj_id[84370] -0.129 0.051 -0.222 -0.031 0.001 0.001 \n", + "Coherence|subj_id_sigma 0.090 0.016 0.063 0.122 0.000 0.000 \n", + "Intercept 6.977 0.064 6.857 7.093 0.003 0.002 \n", + "log_RTs_sigma 0.493 0.003 0.488 0.498 0.000 0.000 \n", + "\n", + " ess_bulk ess_tail r_hat \n", + "1|subj_id[31727] 611.0 974.0 1.01 \n", + "1|subj_id[71329] 574.0 917.0 1.01 \n", + "1|subj_id[71737] 588.0 1095.0 1.01 \n", + "1|subj_id[75445] 633.0 1069.0 1.01 \n", + "1|subj_id[77704] 581.0 1041.0 1.01 \n", + "... ... ... ... \n", + "Coherence|subj_id[84322] 2958.0 2790.0 1.00 \n", + "Coherence|subj_id[84370] 4995.0 2613.0 1.00 \n", + "Coherence|subj_id_sigma 1571.0 2345.0 1.00 \n", + "Intercept 508.0 737.0 1.01 \n", + "log_RTs_sigma 6013.0 2276.0 1.00 \n", + "\n", + "[65 rows x 9 columns]" ] }, - "execution_count": 92, + "execution_count": 6, "metadata": {}, "output_type": "execute_result" } ], + "source": [ + "var_both_para = az.summary(var_both_trace)\n", + "var_both_para" + ] + }, + { + "cell_type": "code", + "execution_count": 15, + "metadata": { + "collapsed": false, + "id": "FECE43D0D5DC45BAA92C4720767C7EDC", + "jupyter": {}, + "notebookId": "676d00454d522334965beb55", + "scrolled": false, + "slideshow": { + "slide_type": "slide" + }, + "tags": [], + "trusted": true + }, + "outputs": [ + { + "data": { + "text/html": [ + "" + ], + "text/plain": [ + "
" + ] + }, + "metadata": {}, + "output_type": "display_data" + } + ], + "source": [ + "# 设置绘图坐标\n", + "figs, (ax1, ax2) = plt.subplots(1,2, figsize = (20,5))\n", + "# 绘制变化的截距\n", + "az.plot_forest(var_both_trace,\n", + " var_names=[\"Coherence\\\\|subj_id\", \"~sigma\", \"~1|\", \"~Intercept\"],\n", + " filter_vars=\"like\",\n", + " combined = True,\n", + " ax=ax1)\n", + "# 绘制变化的斜率\n", + "az.plot_forest(var_both_trace,\n", + " var_names=[\"1|subj_id\", \"~sigma\", \"~Coherence\", \"~Intercept\"],\n", + " filter_vars=\"like\",\n", + " combined = True,\n", + " ax=ax2)\n", + "plt.show()" + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "946E34AD2A77416C88E5E94EA3B37A33", + "jupyter": {}, + "notebookId": "676d00454d522334965beb55", + "runtime": { + "execution_status": null, + "is_visible": false, + "status": "default" + }, + "scrolled": false, + "slideshow": { + "slide_type": "slide" + }, + "tags": [] + }, + "source": [ + "### 模型比较 \n", + "\n", + "##### 模型评估指标 \n", + "\n", + "在分析模型的预测能力时，有绝对指标和相对指标，绝对指标用于衡量模型预测值与真实值之间的差异，相对指标用于比较不同模型的预测能力，通常用于不同方法或模型之间的性能对比。 \n", + "\n", + "##### 绝对指标： \n", + "\n", + "* 在之前的课程中介绍过对后验预测结果进行评估的两种方法 \n", + "\n", + "* 一是**MAE**，即后验预测值与真实值之间预测误差的中位数，二是**within_95**，即真实值是否落在95%后验预测区间内 \n", + "\n", + "* 在这里调用之前写过的计算两种指标的方法，评估两种分层模型的后验预测结果 \n", + "\n", + "\n", + "##### 相对指标 \n", + "\n", + "在实际操作中，我们通过 `ArViz` 的函数`az.loo`计算 $ELPD_{LOO-CV}$。 \n", + "\n", + "PSIS-LOO-CV 有两大优势： \n", + "1. 计算速度快，且结果稳健 \n", + "2. 提供了丰富的模型诊断指标 " + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": { + "collapsed": false, + "id": "692BBB5B977743379984A197FADC464B", + "jupyter": {}, + "notebookId": "676d00454d522334965beb55", + "scrolled": false, + "slideshow": { + "slide_type": "slide" + }, + "tags": [], + "trusted": true + }, + "outputs": [], + "source": [ + "complete_pooled_model.predict(complete_pooled_trace, kind=\"pps\")\n", + "var_inter_model.predict(var_inter_trace, kind=\"pps\")\n", + "var_both_model.predict(var_both_trace, kind=\"pps\")" + ] + }, + { + "cell_type": "code", + "execution_count": 17, + "metadata": { + "collapsed": false, + "id": "B19FDE4283A044CE8D6DF836C3B6EE6A", + "jupyter": {}, + "notebookId": "676d00454d522334965beb55", + "scrolled": false, + "slideshow": { + "slide_type": "slide" + }, + "tags": [], + "trusted": true + }, + "outputs": [], "source": [ "complete_pooled_trace" ] }, { "cell_type": "code", - "execution_count": 93, - "metadata": {}, + "execution_count": 18, + "metadata": { + "collapsed": false, + "id": "460F204ACC2A4BF39C0CE5B6D4A2C8F5", + "jupyter": {}, + "notebookId": "676d00454d522334965beb55", + "scrolled": false, + "slideshow": { + "slide_type": "slide" + }, + "tags": [], + "trusted": true + }, "outputs": [], "source": [ "# 定义计算 MAE 函数\n", @@ -3992,8 +2111,19 @@ }, { "cell_type": "code", - "execution_count": 102, - "metadata": {}, + "execution_count": 19, + "metadata": { + "collapsed": false, + "id": "EBFF2584F1A74483A73192C4965E7DBA", + "jupyter": {}, + "notebookId": "676d00454d522334965beb55", + "scrolled": false, + "slideshow": { + "slide_type": "slide" + }, + "tags": [], + "trusted": true + }, "outputs": [], "source": [ "# 定义\n", @@ -4016,8 +2146,19 @@ }, { "cell_type": "code", - "execution_count": 103, - "metadata": {}, + "execution_count": 20, + "metadata": { + "collapsed": false, + "id": "A5479E2F9C9E48578537B08209479BBF", + "jupyter": {}, + "notebookId": "676d00454d522334965beb55", + "scrolled": false, + "slideshow": { + "slide_type": "slide" + }, + "tags": [], + "trusted": true + }, "outputs": [ { "data": { @@ -4049,20 +2190,20 @@ " \n", " 0\n", " 完全池化\n", - " 0.517794\n", - " 125\n", + " 0.349905\n", + " 910\n", " \n", " \n", " 1\n", " 变化截距\n", - " 0.316176\n", - " 158\n", + " 0.299413\n", + " 936\n", " \n", " \n", " 2\n", " 变化截距、斜率\n", - " 0.307307\n", - " 154\n", + " 0.298617\n", + " 926\n", " \n", " \n", "\n", @@ -4070,12 +2211,12 @@ ], "text/plain": [ " Model MAE Outliers\n", - "0 完全池化 0.517794 125\n", - "1 变化截距 0.316176 158\n", - "2 变化截距、斜率 0.307307 154" + "0 完全池化 0.349905 910\n", + "1 变化截距 0.299413 936\n", + "2 变化截距、斜率 0.298617 926" ] }, - "execution_count": 103, + "execution_count": 20, "metadata": {}, "output_type": "execute_result" } @@ -4102,8 +2243,19 @@ }, { "cell_type": "code", - "execution_count": 104, - "metadata": {}, + "execution_count": 21, + "metadata": { + "collapsed": false, + "id": "437531310A0E4D79B0814FEAACE14F03", + "jupyter": {}, + "notebookId": "676d00454d522334965beb55", + "scrolled": false, + "slideshow": { + "slide_type": "slide" + }, + "tags": [], + "trusted": true + }, "outputs": [ { "data": { @@ -4141,11 +2293,11 @@ " \n", " model3(hierarchy both)\n", " 0\n", - " -1941.431752\n", - " 13.064896\n", + " -11583.298448\n", + " 60.429914\n", " 0.000000\n", - " 0.884853\n", - " 37.106601\n", + " 0.930945\n", + " 117.897744\n", " 0.000000\n", " False\n", " log\n", @@ -4153,24 +2305,24 @@ " \n", " model2(hierarchical intercept)\n", " 1\n", - " -1950.120395\n", - " 8.071729\n", - " 8.688643\n", - " 0.115147\n", - " 36.892828\n", - " 4.741897\n", + " -11622.610522\n", + " 34.771387\n", + " 39.312074\n", + " 0.000000\n", + " 117.434555\n", + " 9.837238\n", " False\n", " log\n", " \n", " \n", " model1(complete pooling)\n", " 2\n", - " -2502.072721\n", - " 3.013059\n", - " 560.640969\n", - " 0.000000\n", - " 34.205582\n", - " 27.353555\n", + " -13692.877664\n", + " 3.690037\n", + " 2109.579216\n", + " 0.069055\n", + " 113.467901\n", + " 72.296478\n", " False\n", " log\n", " \n", @@ -4179,18 +2331,18 @@ "" ], "text/plain": [ - " rank elpd_loo p_loo elpd_diff \\\n", - "model3(hierarchy both) 0 -1941.431752 13.064896 0.000000 \n", - "model2(hierarchical intercept) 1 -1950.120395 8.071729 8.688643 \n", - "model1(complete pooling) 2 -2502.072721 3.013059 560.640969 \n", - "\n", - " weight se dse warning scale \n", - "model3(hierarchy both) 0.884853 37.106601 0.000000 False log \n", - "model2(hierarchical intercept) 0.115147 36.892828 4.741897 False log \n", - "model1(complete pooling) 0.000000 34.205582 27.353555 False log " + " rank elpd_loo p_loo elpd_diff \\\n", + "model3(hierarchy both) 0 -11583.298448 60.429914 0.000000 \n", + "model2(hierarchical intercept) 1 -11622.610522 34.771387 39.312074 \n", + "model1(complete pooling) 2 -13692.877664 3.690037 2109.579216 \n", + "\n", + " weight se dse warning scale \n", + "model3(hierarchy both) 0.930945 117.897744 0.000000 False log \n", + "model2(hierarchical intercept) 0.000000 117.434555 9.837238 False log \n", + "model1(complete pooling) 0.069055 113.467901 72.296478 False log " ] }, - "execution_count": 104, + "execution_count": 21, "metadata": {}, "output_type": "execute_result" } @@ -4206,9 +2358,23 @@ }, { "cell_type": "markdown", - "metadata": {}, + "metadata": { + "id": "B336E89438EB443FB0704E77C8BF86A3", + "jupyter": {}, + "notebookId": "676d00454d522334965beb55", + "runtime": { + "execution_status": null, + "is_visible": false, + "status": "default" + }, + "scrolled": false, + "slideshow": { + "slide_type": "slide" + }, + "tags": [] + }, "source": [ - "通过 `arviz.compare` 方法来对比多个模型的 elpd。从下面结果可见： \n", + "通过 `arviz.compare` 方法来对比多个模型的 elpd。从下面结果可见： \n", " \n", "- 模型3的 elpd_loo 最大，表明它对**样本外数据**的预测性能最好。 \n", "- 而模型1的 elpd_loo 最小，表明它的预测性能最差。 \n", @@ -4218,35 +2384,58 @@ }, { "cell_type": "markdown", - "metadata": {}, + "metadata": { + "id": "044339620F5744FFBEE390917464111A", + "jupyter": {}, + "notebookId": "676d00454d522334965beb55", + "runtime": { + "execution_status": null, + "is_visible": false, + "status": "default" + }, + "scrolled": false, + "slideshow": { + "slide_type": "slide" + }, + "tags": [] + }, "source": [ - "### 贝叶斯统计推断\n", - "\n", - "模型比较发现，模型3（变化截距和变化斜率）的结果在3个模型中是最好的。最后，将使用 HDI + ROPE 和贝叶斯因子（Bayes Factor，BF）来进行统计推断。\n", + "### 贝叶斯统计推断 \n", "\n", - "#### HDI + ROPE 的统计推断\n", + "模型比较发现，模型3（变化截距和变化斜率）的结果在3个模型中是最好的。最后，将使用 HDI + ROPE 和贝叶斯因子（Bayes Factor，BF）来进行统计推断。 \n", "\n", - "* 反应时差异的后验分布平均值为 -281 ms，然而，这一数值并不足以断定实验条件对反应时间有显著的减少作用。\n", - "* 95% HDI 范围为 [-608 ms, 48 ms]，表明后验分布中95%的概率下的反应时差异位于此区间。由于95% HDI 包含了0，并且分布主要集中在负值方向，但这一趋势并不足以证明存在显著的效应。\n", - "* ROPE 设定了一个 [-30 ms, 30 ms] 的实用等效区间，用以判断反应时差异是否具有实际意义。ROPE 内的概率仅为1.6%，尽管这表明在大多数情况下反应时差异超出了可忽略的范围，但这一差异仍不足以被视为显著。\n", + "#### HDI + ROPE 的统计推断 \n", "\n", + "* 反应时差异的后验分布平均值为 -174 ms，然而，这一数值并不足以断定实验条件对反应时间有显著的减少作用。 \n", + "* 95% HDI 范围为 [-231 ms, -121 ms]，表明后验分布中95%的概率下的反应时差异位于此区间。由于95% HDI 不包含0，并且分布主要集中在负值方向，表明实验条件对反应时间存在显著效应的。 \n", + "* ROPE 设定了一个 [-30 ms, 30 ms] 的实用等效区间，用以判断反应时差异是否具有实际意义。图中显示“0.0% in ROPE”，表明估计的差异没有落在ROPE范围内，这意味着反应时间差异是统计上显著的。 \n", "\n", - "#### 贝叶斯因子（Bayes Factor，BF）\n", + "#### 贝叶斯因子（Bayes Factor，BF） \n", "\n", - "反应时的差异（beta_1）在统计上和实际意义上均不显著。\n", - "* 数据强烈支持 无效假设（beta_1 = 0），即反应时差异可能不存在或非常微弱。\n", - "* 贝叶斯因子 BF_10 = 0.01 提供了明确的证据，表明 beta_1 不显著。\n", - "* 从后验分布来看，数据更新后 beta_1 的可能值仍然集中在 0 附近，进一步支持无效假设。\n" + "* 贝叶斯因子 BF_10 = 17.32， BF_01 = 0.06，表明数据有较强的证据支持备择假设。" ] }, { "cell_type": "code", - "execution_count": 106, - "metadata": {}, + "execution_count": 11, + "metadata": { + "collapsed": false, + "id": "C6FAF83B9D754789AFED826AAAB5C19C", + "jupyter": {}, + "notebookId": "676d00454d522334965beb55", + "scrolled": false, + "slideshow": { + "slide_type": "slide" + }, + "tags": [], + "trusted": true + }, "outputs": [ { "data": { - "image/png": "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", + "text/html": [ + "" + ], "text/plain": [ "
" ] @@ -4291,19 +2480,127 @@ }, { "cell_type": "code", - "execution_count": 108, - "metadata": {}, + "execution_count": 30, + "metadata": { + "collapsed": false, + "id": "281E29CFA371463D98CB1ABDF6E23B6B", + "jupyter": {}, + "notebookId": "676d00454d522334965beb55", + "scrolled": false, + "slideshow": { + "slide_type": "slide" + }, + "tags": [], + "trusted": true + }, + "outputs": [], + "source": [ + "import arviz as az\n", + "\n", + "# 假设 var_both_trace 是一个 InferenceData 对象\n", + "# 删除 prior 和 prior_predictive 数据\n", + "if hasattr(var_both_trace, \"prior\"):\n", + " del var_both_trace.prior\n", + "\n", + "if hasattr(var_both_trace, \"prior_predictive\"):\n", + " del var_both_trace.prior_predictive\n", + "\n", + "# 检查删除后的 InferenceData 对象\n", + "print(var_both_trace)" + ] + }, + { + "cell_type": "code", + "execution_count": 33, + "metadata": { + "collapsed": false, + "id": "7DB99E80FB4244BF904F9C6BFE6D75F3", + "jupyter": {}, + "notebookId": "676d00454d522334965beb55", + "scrolled": false, + "slideshow": { + "slide_type": "slide" + }, + "tags": [], + "trusted": true + }, + "outputs": [], + "source": [ + "var_both_trace" + ] + }, + { + "cell_type": "code", + "execution_count": 10, + "metadata": { + "collapsed": false, + "id": "B73FB7BF12214713A29AD8F6D25C8B31", + "jupyter": {}, + "notebookId": "676d00454d522334965beb55", + "scrolled": false, + "slideshow": { + "slide_type": "slide" + }, + "tags": [], + "trusted": true + }, + "outputs": [ + { + "data": { + "text/plain": [ + "" + ] + }, + "execution_count": 10, + "metadata": {}, + "output_type": "execute_result" + }, + { + "data": { + "text/html": [ + "" + ], + "text/plain": [ + "
" + ] + }, + "metadata": {}, + "output_type": "display_data" + } + ], + "source": [ + "az.plot_posterior(var_both_trace.prior[\"Coherence\"])" + ] + }, + { + "cell_type": "code", + "execution_count": 8, + "metadata": { + "collapsed": false, + "id": "8B9D990D155448FABD596881A94A0701", + "jupyter": {}, + "notebookId": "676d00454d522334965beb55", + "scrolled": false, + "slideshow": { + "slide_type": "slide" + }, + "tags": [], + "trusted": true + }, "outputs": [ { "name": "stderr", "output_type": "stream", "text": [ - "Sampling: [1|subj_id_offset, 1|subj_id_sigma, Coherence, Coherence|subj_id_offset, Coherence|subj_id_sigma, Intercept, log_RTs, log_RTs_sigma]\n" + "Sampling: [1|subj_id_offset, 1|subj_id_sigma, Coherence, Coherence|subj_id_offset, Coherence|subj_id_sigma, Intercept, log_RTs, log_RTs_sigma]\n", + "The reference value is outside of the posterior. This translate into infinite support for H1, which is most likely an overstatement.\n" ] }, { "data": { - "image/png": "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", + "text/html": [ + "" + ], "text/plain": [ "
" ] @@ -4314,7 +2611,7 @@ ], "source": [ "# 进行贝叶斯因子计算，需要采样先验分布\n", - "var_both_trace.extend(var_both_model.prior_predictive(random_seed=84735) )\n", + "var_both_trace.extend(var_both_model.prior_predictive(draws=5000, random_seed=84735) )\n", "\n", "# 绘制贝叶斯因子图\n", "az.plot_bf(var_both_trace, var_name=\"Coherence\", ref_val=0)\n", @@ -4328,48 +2625,76 @@ }, { "cell_type": "markdown", - "metadata": {}, + "metadata": { + "id": "AAD83F64719B493192A71A819FD95BFC", + "jupyter": {}, + "notebookId": "676d00454d522334965beb55", + "runtime": { + "execution_status": null, + "is_visible": false, + "status": "default" + }, + "scrolled": false, + "slideshow": { + "slide_type": "slide" + }, + "tags": [] + }, "source": [ - "### 结论\n", + "### 结论 \n", "\n", "\n", - "通过模型建立和模型比较，相比于模型1和模型2，模型3（变化截距和变化斜率）的效果最好。因此，在随机点运动范式中，反应时间显著受到随机点运动方向一致性比例的影响。\n", + "通过模型建立和模型比较，相比于模型1和模型2，模型3（变化截距和变化斜率）的效果最好。因此，在随机点运动范式中，反应时间显著受到随机点运动方向一致性比例的影响。 \n", "\n", - "具体而言，模型3考虑了变化截距和变化斜率，这意味着它不仅捕捉到了不同条件下反应时间的平均差异，还考虑了这些差异随条件变化的趋势。相比之下，模型1和模型2未能充分捕捉到这些复杂的变化模式，因而在预测精度和模型拟合度上表现不如模型3。\n" + "具体而言，模型3考虑了变化截距和变化斜率，这意味着它不仅捕捉到了不同条件下反应时间的平均差异，还考虑了这些差异随条件变化的趋势。相比之下，模型1和模型2未能充分捕捉到这些复杂的变化模式，因而在预测精度和模型拟合度上表现不如模型3。 \n" ] }, { "cell_type": "markdown", - "metadata": {}, + "metadata": { + "id": "F7230000F8BA4FCA9FBD74E0170017C8", + "jupyter": {}, + "notebookId": "676d00454d522334965beb55", + "runtime": { + "execution_status": null, + "is_visible": false, + "status": "default" + }, + "scrolled": false, + "slideshow": { + "slide_type": "slide" + }, + "tags": [] + }, "source": [ - "### 大作业注意事项\n", + "### 大作业注意事项 \n", "\n", - "在完成大作业时，请注意以下几点要求：\n", + "在完成大作业时，请注意以下几点要求： \n", "\n", - "1. **文档**：\n", - " - 请按照APA7论文格式撰写文档，确保内容完整、格式规范。\n", - " - 文档应包括研究背景、方法、结果和讨论等部分，详细描述研究过程和发现。\n", + "1. **文档**： \n", + " - 请按照APA7论文格式撰写文档，确保内容完整、格式规范。 \n", + " - 文档应包括研究背景、方法、结果和讨论等部分，详细描述研究过程和发现。 \n", "\n", - "2. **和鲸Notebook演示（或PPT）**：\n", - " - 使用和鲸Notebook或PPT进行演示，清晰展示研究的各个步骤和结果。\n", - " - 演示内容应包括数据处理、模型构建、结果分析和结论等部分。\n", + "2. **和鲸Notebook演示（或PPT）**： \n", + " - 使用和鲸Notebook或PPT进行演示，清晰展示研究的各个步骤和结果。 \n", + " - 演示内容应包括数据处理、模型构建、结果分析和结论等部分。 \n", "\n", - "3. **代码**：\n", - " - 提交完整的代码，确保代码可以运行并生成预期结果。\n", - " - 代码应包括数据导入、预处理、模型构建、拟合和结果分析等部分。\n", - " - 请在代码中添加必要的注释。\n", + "3. **代码**： \n", + " - 提交完整的代码，确保代码可以运行并生成预期结果。 \n", + " - 代码应包括数据导入、预处理、模型构建、拟合和结果分析等部分。 \n", + " - 请在代码中添加必要的注释。 \n", "\n", "\n", - "💡互评环节：\n", + "💡互评环节： \n", "\n", - "在2025年1月3号的最后一次课时，各小组将进行大作业汇报。每组可以对其他汇报组进行评分，评分标准包括文档规范性、演示效果和代码运行情况等。\n", + "在2025年1月3号的最后一次课时，各小组将进行大作业汇报。每组可以对其他汇报组进行评分，评分标准包括文档规范性、演示效果和代码运行情况等。 \n", "\n" ] } ], "metadata": { "kernelspec": { - "display_name": "pymc5_3.11", + "display_name": "Python 3", "language": "python", "name": "python3" }, @@ -4383,7 +2708,7 @@ "name": "python", "nbconvert_exporter": "python", "pygments_lexer": "ipython3", - "version": "3.11.5" + "version": "3.5.2" } }, "nbformat": 4,