Example of using Tea

tea_example_0.1.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "Example of using Tea",
      "provenance": [],
      "collapsed_sections": [],
      "include_colab_link": true
    },
    "kernelspec": {
      "display_name": "Python 3",
      "name": "python3"
    }
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/gist/emjun/cca25cdbeee39afe90771dde9cb80457/tea_example_0.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "colab_type": "text",
        "id": "JyG45Qk3qQLS"
      },
      "source": [
        "# Tea\n",
        "Tea requires users to describe their data, variables, study design, assumptions about the data, and hypotheses at a high-level. Tea combines computed properties about the data (e.g., normal distribution) with users' assumptions and hypotheses to infer a set of valid statisitcal analyses that test users' hypotheses. Unlike other statistical analysis tools, Tea focuses on capturing users' *explicit* hypotheses and assumptions about the data and does *not* require users to specify the specific statistical tests. \n",
        "\n",
        "Tea is designed for non-statistical experts!!"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Ua--kwl8Ge7x",
        "colab_type": "text"
      },
      "source": [
        "## Example\n",
        "\n",
        "Let's walk through an example! Make sure to [install Tea before](http://tea-lang.org/install).  :)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "VLt-_g2lGtHi",
        "colab_type": "text"
      },
      "source": [
        "## 1. Import tea"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "-gsAYPFgGxhL",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "import tea"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "colab_type": "text",
        "id": "KR921S_OQSHG"
      },
      "source": [
        "## Data\n",
        "\n",
        "**Load data.**\n",
        "\n",
        "This example is taken from Ehrlich[1] and Vandaele[2].  The data set comes as part of the MASS package in R.\n",
        "\n",
        "Let's say you're a historical criminologist who wants to know \"Is there a significant difference in imprisonment probabilities between Southern and non-Southern states?\"\n",
        "\n",
        "\n",
        "\n",
        "[1] Isaac Ehrlich. 1973. Participation in illegitimate activities: A theoretical and empirical investigation. Journal of political Economy. 81, 3 (1973), 521–565.\n",
        "\n",
        "[2] Walter Vandaele. 1987. Participation in illegitimate activities: Ehrlich revisited, 1960. Vol. 8677. Inter-university Consortium for Political and Social Research."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "cellView": "both",
        "colab_type": "code",
        "id": "WUtu4316QSHL",
        "colab": {}
      },
      "source": [
        "\n",
        "tea.data(\"https://homes.cs.washington.edu/~emjun/tea-lang/datasets/UScrime.csv\")\n"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "colab_type": "text",
        "id": "Id6tDF1HQSHD"
      },
      "source": [
        "## Variables\n",
        "\n",
        "**Declare and annotate the variables of interest.**\n",
        "\n",
        "There are two variables: `So` and `Prob`. \n",
        "\n",
        "`So` is a binary nominal variable where `So=1` means the state is *Southern* and `So=0` means the state is *non-Southern*.\n",
        "\n",
        "`Prob` is a ratio variable for the probability of imprisonment in each state."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "9_3Uxu7cGS0D",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "variables = [\n",
        "    {\n",
        "        'name' : 'So',\n",
        "        'data type' : 'nominal',   # Options: 'nominal', 'ordinal', 'interval', 'ratio'\n",
        "        'categories' : ['0', '1']\n",
        "    },\n",
        "    {\n",
        "        'name' : 'Prob',\n",
        "        'data type' : 'ratio',   # Options: 'nominal', 'ordinal', 'interval', 'ratio'\n",
        "        'range' : [0,1]   # optional\n",
        "    }\n",
        "]\n",
        "\n",
        "tea.define_variables(variables)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "qwy7tpFCHlkw",
        "colab_type": "text"
      },
      "source": [
        "## Assumptions\n",
        "\n",
        "**OPTIONAL: Declare any assumptions you may have about the data based on prior visualization or domain knowledge.**\n",
        "\n",
        "Based on prior knowledge in historical criminology, you might assume that probability of imprisonment is normally distributed in Southern and non-Southern states (the two groups of interest).\n",
        "\n",
        "If no Type I Error Rate (or \"significance threshold\") is specified, Tea will use .05."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "JIrQI3r4NZRI",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "assumptions = {\n",
        "    'groups normally distributed': [['So', 'Prob']],\n",
        "    'Type I (False Positive) Error Rate': 0.05,\n",
        "}\n",
        "\n",
        "tea.assume(assumptions)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "1fCGhA2DNbHh",
        "colab_type": "text"
      },
      "source": [
        "## Study Design\n",
        "\n",
        "**Express how the data were collected.**"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "DiZp2K-rN24X",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "experimental_design = {\n",
        "                        'study type': 'observational study',   # 'study type' could be 'experiment'\n",
        "                        'contributor variables': 'So',   # 'experiment's have 'independent variables'\n",
        "                        'outcome variables': 'Prob',   # 'experiment's have 'dependent variables'\n",
        "                    }\n",
        "tea.define_study_design(experimental_design)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "lFc-d10JN3_i",
        "colab_type": "text"
      },
      "source": [
        "## Hypothesis\n",
        "**Explicitly state a hypothesis about the relationship between the variables in the data.**\n",
        "\n",
        "Based on your domain knowledge, you might hypothesize that there is a relationship between a state being Southern/non-Southern (`So`) and the probability of imprisonment (`Prob`).\n",
        "In particular, you might hypothesize that Southern states (`So = 1`) have higher (`>`) imprisonment probabilities than non-Southern states (`So = 0`)."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "5mNM6Q8uOE3h",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "tea.hypothesize(['So', 'Prob'], ['So:1 > 0'])"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "NdBJgR3hYc3Q",
        "colab_type": "text"
      },
      "source": [
        "# Hope that was as fun and easy as a cup of tea! ;)\n",
        "If you have any **COMMENTS OR FEEDBACK** please do not hesitate to get in touch: emjun [at] cs dot washington dot edu or join us in the [Tea Room](https://gitter.im/tea-room)."
      ]
    }
  ]
}