Data_Set_Description.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Dataset Documentation\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Observational Study\n",
    "\n",
    "\n",
    "#### 1. Motivation of data collection \n",
    "The data have been scraped from Instagram for academic purposes (July 2020). The authors  created and prepared the dataset for analysis.\n",
    "\n",
    "\n",
    "#### 2. Composition of dataset \n",
    "The dataset provides a list of 111960 historical posts from 316 Instagram users. The `consumers_psm` table can be used to select usernames included in the matched sample. Image content associated with these posts have been analyzed with Azure Vision API to compute image-similarity within and between-subjects. All data have been normalized and separated into various schemas, whose variable operationalisations we describe next.\n",
    "\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### `consumers_country`\n",
    "\n",
    "| ** Variable ** | ** Operationalisation ** | \n",
    "| :---- | :--- | \n",
    "| username | Instagram username of the account |\n",
    "| country | The country of origin of an Instagram user                                                  |"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### `consumers_posts_json`\n",
    "\n",
    "| ** Variable ** | ** Operationalisation ** | \n",
    "| :---- | :--- | \n",
    "| username | Instagram username of the account |\n",
    "| date | The date of data retrieval (i.e., scraping from Instagram) |\n",
    "| posts | A dictionary that contains individual level posts from a username (formatted in JSON) |\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### `consumers_posts`\n",
    "\n",
    "| ** Variable ** | ** Operationalisation ** | \n",
    "| :---- | :--- | \n",
    "| content_type | Type of content, GraphImage (images), GraphSidecar (carousel), or GraphVideo (video) |\n",
    "| media1 | Link to media file (URL signature only valid for limited time) |\n",
    "| shortcode | Instagram post ID (visit instagram.com/p/{shortcode} to view the associated post)|\n",
    "| total_comments | Number of comments at data retrieval |\n",
    "| username | Instagram username of the account |\n",
    "| video_views | Number of video views at data retrieval (null for images) |\n",
    "| date | Publication date of the post                                                             |"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### `consumers_profile`\n",
    "\n",
    "| ** Variable ** | ** Operationalisation ** | \n",
    "| :---- | :--- | \n",
    "| biography | A section on a user's profile page to include personal  information |\n",
    "| followers_count | The number of accounts that follow the user at data retrieval |\n",
    "| following_count | The number of accounts a user follows at data retrieval|\n",
    "| full_name | The first and last name of a user (may deviate from the username) |\n",
    "| posts_count | The total number of posts on Instagram for a user at data retrieval|\n",
    "| username | Instagram username of the account                                                 |\n",
    "\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### `consumers_profile_json`\n",
    "\n",
    "| ** Variable ** | ** Operationalisation ** | \n",
    "| :---- | :--- | \n",
    "| username | Instagram username of the account |\n",
    "| date | The date of data retrieval (i.e., scraping from Instagram) |\n",
    "| profile | A dictionary that contains Instagram profile information from a username (formatted in JSON)                         |"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### `consumers_psm`\n",
    "\n",
    "| ** Variable ** | ** Operationalisation ** | \n",
    "| :---- | :--- | \n",
    "| username | Instagram username of the account (only includes usernames that are a part of the matched sample) |\n",
    "| type | Indicator variable for the test condition (either treatment or control)                                     |"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### `hypeauditor`\n",
    "\n",
    "| ** Variable ** | ** Operationalisation ** | \n",
    "| :---- | :--- | \n",
    "| followers_from_country | The number of accounts that follow the user from a given country at data retrieval (i.e., subset of total_follower)|\n",
    "| topics | Tag words associated with the image content of the user|\n",
    "| total_follower| The total number of accounts that follow the user at data retrieval |\n",
    "| url| Link to HypeAuditor ranking for a topic (country can be extracted from the URL)|\n",
    "| username| Instagram username of the influencer account                                                |"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### `influencer_country_purity`\n",
    "\n",
    "| ** Variable ** | ** Operationalisation ** | \n",
    "| :---- | :--- | \n",
    "| country | The major country of origin among the influencer's followers  |\n",
    "| percentage_country | The percentage of the influencer's followers located in `country` |\n",
    "| username| Instagram username of the influencer account |"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": true
   },
   "source": [
    "#### `image_similarity_between_tags`\n",
    "\n",
    "| ** Variable ** | ** Operationalisation ** | \n",
    "| :---- | :--- | \n",
    "| username1 | Instagram username of the account whose image content is compared with username2 |\n",
    "| username2 | Instagram username of the account whose image content is compared with username1 |\n",
    "| username1_2 | Concatenation of username1 and username2 in alphabetical order to identify overlapping pairs: (username1, username2) and (username2, username1) |\n",
    "| before_after | Indicator variable for the intervention time frame (either before or after) |\n",
    "| similarity | Mean cosine similarity of images of username1 and username2 before or after hiding like counts                                                             |"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### `image_similarity_within_tags`\n",
    "\n",
    "| ** Variable ** | ** Operationalisation ** | \n",
    "| :---- | :--- | \n",
    "| username | Instagram username of the account |\n",
    "| before_after | Indicator variable for the intervention time frame (either before or after) |\n",
    "| image_similarity | Mean cosine similarity of all images of a user before or after hiding like counts                                                 |"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 3. Collection Process\n",
    "The data was collected in July 2020 and consists of posts ranging between 2011 and 2020 (depending on when users joined Instagram). We used an [external library](https://github.com/arc298/instagram-scraper) to scrape posts from selected Instagram accounts. Our start seed of influencer accounts was obtained in February 2020 from [HypeAuditor](https://hypeauditor.com/top-instagram/). The dataset only comprises data of users whose Instagram account is public (no private data). Furthermore, image content data was obtained through [Azure Computer Vision API](https://azure.microsoft.com/en-us/services/cognitive-services/computer-vision/).\n",
    "\n",
    "#### 4. Preprocessing\n",
    "A detailed description of the preprocessing of raw JSON data can be found over [here](https://github.com/RoyKlaasseBos/Hiding-Instagram-Likes/blob/master/Web_Appendix_Hiding_Like_Counts_Instagram.ipynb). The data has been cleaned using a combination of Python and R.\n",
    "\n",
    "#### 5. Uses\n",
    "The dataset has been used to study the effect of hiding like counts on user behavior and self-esteem. To this end, the data consists of observational data before and after the intervention. Other use cases for the dataset are natural language processing and object recognition (deep learning).\n",
    "\n",
    "#### 6. Distribution\n",
    "The repository containing all code is publicly available on Github. Raw and preprocessed data is stored in a cloud database. Anonymized data may be requested for academic purposes only.\n",
    "\n",
    "#### 7. Maintenance\n",
    "The dataset is stored on Amazon Web Services (RDS) and backed-up daily."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Experiment"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### `experiment`\n",
    "\n",
    "| ** Variable ** | ** Operationalisation ** | \n",
    "| :---- | :--- | \n",
    "| gender | The gender of a participant (1=Male, 2=Female, 3=Other)|\n",
    "| condition | The test condition a participant was assigned to (low, high, or hidden likes)|\n",
    "| gender_condition | The combination of the gender of the person shown on the image and the like count test condition ({female, male} X {low, high, hidden})|\n",
    "| other.evaluation_1 | The extent to which the target person is considered likeable (1=Strongly disagree, 7=Strongly agree) |\n",
    "| other.evaluation_2 | The extent to which the target person is considered popular (1=Strongly disagree, 7=Strongly agree) |\n",
    "| other.evaluation_3 | The extent to which the target person is considered attractive (1=Strongly disagree, 7=Strongly agree) |\n",
    "| self.evaluation_1 | The extent to which the participant considers himself/herself likeable (1=Strongly disagree, 7=Strongly agree) |\n",
    "| self.evaluation_2 | The extent to which the participant considers himself/herself popular (1=Strongly disagree, 7=Strongly agree) |\n",
    "| self.evaluation_3 | The extent to which the participant considers himself/herself attractive (1=Strongly disagree, 7=Strongly agree) |\n",
    "| instagram_actions_1 | The participant's likelihood to like the photo (1=Extremely unlikely, 7=Extremely likely)|\n",
    "| instagram_actions_2 | The participant's likelihood to comment on the photo (1=Extremely unlikely, 7=Extremely likely)|\n",
    "| instagram_actions_3 | The participant's likelihood to follow the Instagram user (1=Extremely unlikely, 7=Extremely likely)|\n",
    "| instagram_actions_4 | The participant's likelihood to share a photo on Instagram (1=Extremely unlikely, 7=Extremely likely)|\n",
    "| instagram_usage | The average time daily spent on Instagram (1=Less than 10 minutes, 2=11-30 minutes, 3=31-60 minutes, 4=1-2 hours, 5=2-3 hours, 6=More than 3 hours).|\n",
    "| age | The participant's age (in years). The survey was conducted among 18-30 year-olds.|\n",
    "| ethnicity | The participant's ethnicity (1=White, 2=Black, 3=Hispanic, 4=Asian, 5=Other)|\n",
    "\n",
    "\n",
    "\n",
    "\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "---"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "*Klaasse Bos, R.J. (2020). Dataset Documentation: Goodbye Likes, Hello Mental Health: How Hiding Like Counts Affects User Behavior & Self-Esteem.*"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.10"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}