COGS9_8_EDA
Page 1: Introduction
Course Title: Exploratory Data Analysis (COGS 9)
Instructor: Dominic Burrows, PhD
Institution: UCSD Cog Sci
Contact Information: dburrows@ucsd.edu | Twitter: @dominicburrows_
Page 2: Course Announcements
Quiz 3: Due Friday at 11:59 pm.
Assignment 1: Due next Friday, February 7th.
Final Project: Group project; part 1 due February 14th.
Feedback: Anonymous feedback link provided.
Page 3: Cognitive Science Board Event
Event: Networking event on January 31st from 4-6 pm in Price Center East Ballroom.
Highlights: Free food, opportunity to network with alumni.
RSVP: Required for attendance.
Page 4: Class Question Sign Up
Platform: ClassQuestion (lecture Q&A).
Link: Access via provided URL.
Entry Code: RQFKP (must sign in with UCSD email).
Page 5: Learning Objectives
Understand:
The purpose of EDA.
Key steps in EDA process.
Page 6: EDA Process Overview
Key Steps: Start with a question, clean the data, visualize, model, predict, analyze & conclude, share insights.
Process Flow: Visual representation of EDA steps, emphasizing exploration and refinement.
Page 8: Definition of EDA
Exploratory Data Analysis (EDA): Initial phase of data analysis focused on examining data structure, contents, and patterns without directly answering the primary data question.
Purpose: Identify patterns to guide further analysis and model refinement.
Page 9-11: EDA Skills & Components
Key Skills:
Hacking skills
Machine learning understanding
Data processing, mathematical/statistical knowledge
Visualizing Data: Essential for understanding trends and insights.
Pages 12-18: EDA Analogy
Analogy: Exploring British culture before visiting the UK.
Aim: Obtain an authentic experience; gather insights from key locations.
Planning: Identifying relations between locations and preparing a travel plan.
Page 20-21: Importance of EDA
Objective: "Get a feel for the dataset" to avoid errors and refine analysis.
Key Benefits:
Identifies data errors and relationships between features.
Reduces noise and aids in hypothesis refinement.
Page 22-24: EDA Case Study
Example: Study on the effects of seizures on zebrafish by Ahrens et al.
Methods: Visually inspect data, identify seizure time points, inspect distributions.
Pages 25-30: Key EDA Steps
Steps:
Explore data distributions
Look at measures of central tendency and variability
Perform statistical tests
Investigate relationships between features
Understanding Distributions: Can help in refining models and analyzing relationships.
Pages 31-36: Understanding Distributions
Spread of Data: Explains why it's essential to know data distribution types (e.g., Normal, Poisson).
Use Cases: Identifying data characteristics helps in statistical modeling and inference.
Key Theorems: Discusses the implications of the Central Limit Theorem.
Pages 37-42: More on Distributions
Different Distribution Types: Uniform, Exponential, Bernoulli, etc.
Applications: Understanding the distribution helps in choosing statistical tests.
Pages 43-49: Implications of Distribution Understanding
Statistical Tests: Using the appropriate test based on data distribution enhances analysis accuracy.
Modeling for Prediction: Different distributions guide the type of models used in predictions.
Pages 50-53: Modeling for Prediction and Inference
Modeling Types: Linear regression for normal data, logistic regression for categorical data, Poisson regression for count data.
Importance: Accurate modeling is crucial for both prediction and understanding data relationships.
Page 54-56: EDA Conclusion and Key Steps
Key Steps Recap: Emphasizes exploration, central tendency, variability, statistical tests, and relationships.
Class Engagement: Encouragement for participation and discussion.
Pages 57-62: Central Tendency and Variability
Measures of Central Tendency: Mean, Median - Importance in summarizing data.
Variability: Understanding diversity among data points to enhance confidence in sample data predictions.
Pages 63-70: Understanding Variability
Variability: Defines measuring variability and its role in understanding dataset differences.
Confidence Intervals: Provides ranges containing true population mean with a specified degree of confidence.
Pages 71-76: Statistical Tests Importance
Statistical Tests: Emphasizes their role in validating hypotheses and analyzing effects.
Pages 77-92: Refinement of Hypotheses
Hypothesis Refinement: Specific to domains like COVID-19; understanding differences in expression levels by gender.
Conclusions from EDA: Data-driven insights that guide further exploration and hypothesis testing.