• Course website: https://adav-course-2024.netlify.app/
• Each week: 1 lecture, 1 practical
  • Lectures: Tuesdays 9.00 - 10.45 OR 11:00 - 12:45
  • Practicals on Thursdays, group-dependent times
• Readings: https://adav-course-2024.netlify.app/course_info/reading/
• This is a parallel course for UU wide and Economics ADS minor students
  • Lectures are together
  • Practicals are UU / Economics-specific
  • Exam is together

• Practical application of the theory learned during the week
• We will use R and R studio
• For the lab sessions, you have to hand in homework which will be graded pass/fail. Deadline two hours before the labs (at 11am).
  • Lab 3 is a ‘do-it-yourself’ lab due to holiday, the full lab needs to be handed in and will account for weeks’ one pass/fail grade
  • Nothing to submit after the labs.

• Answers for each lab can be viewed on the course website on Friday.

• Individual homeworks: 20% of your grade
  • A composite score of all pass/fails of lab homework + extra question (week 2)
• Group Assignment (in 3 or 4): 20% of your grade
  • Analyzing data and interactive visualization using data science methods R
  • Part 1 of group assignment is optional (but gives you +0.5 points bonus towards your group assignment grade)
• Exam: 60% of your grade
  • Understanding and comprehension of data analysis methods, theory and skills
  • Digital exam in Remindo, exam is on location

• Lectures (apart from week 8) and labs are on location
• Lecture on week 8 will be ONLINE
• Attendance is mandatory
• Locations may be differ over the weeks, please consult your time table at https://mytimetable.uu.nl
• Assignment/homeworks submissions via Blackboard

• You can use AI tools to ask about concepts you didn’t fully understand
• You are NOT allowed to copy paste questions to AI tools. Having AI write your homework/assignment constitutes plagiarism.
• In your assignment, you have to acknowledge and specify how you used any AI tool
• It is important to remember that chatGPT and other AI tools are not a replacement for your own critical thinking and original ideas. The ultimate goal of this course and any tool used to submit work is to enhance your own learning and understanding, not to undermine it.
• If the teachers suspects any violation of this policy, he/she can ask further questions to you on the assignment/homework during the lab session.

• Allowed: What is the difference of logistic regression and linear regression?
• Allowed: What is the function to create a histogram in R?
• Not allowed: Generate all the code for the following assignment.

• What are statistical learning / data analysis / data visualisation
• Why do we need them
• Different types of data analysis
• Key data science concepts
  • Supervised vs unsupervised learning
  • Trade-off between model accuracy and interpretability

• Machine Learning and Recommendations in Netflix

• Netflix Movies and TV Shows dataset from Kaggle

• “Data modeling”?
• “Machine learning”?
• “Statistical learning”?
• “Statistics”?
• “Data science”?
• “Data mining”?

• People from different fields (statistics, computer science, information science, industry) have different goals and standard approaches.
• But they often use the same techniques.
• So each field came up with words that emphasized a different aspect of “data analysis”.
• Because of this history, the terms are not exact synonyms.
• But most people can’t tell the difference!

Our emphasis is going to be on creating results that help you (the analyst) understand the data and make predictions.

• You are statistical consultant hired by a client to investigate the association between advertising and sales
• The asked you to predict sales increase given an increase in advertising budget
• Advertising budget: Input variables (predictors, features, independent variables), denoted as X, X1 the TV, X2 the radio, X3 the newspaper budget
• Sales: Output variable (response, dependent variable) denoted as Y

Oversimplification:

• Data analysis: estimating a model f to summarize our data, composed of an outcome Y and a set of predictors \(X: \hat{Y}=\hat{f(X)}\).
• Data science: mostly interested in prediction, so in \(\hat{Y}\)
• Empirical research / economics: mostly interested in inference (i.e., explanation): understanding the relationship between X and Y
  • (so we are interested in the model that generates the predictions)
• Example: use linear regression to summarize our data \(\hat{y} = α + β1X1 + β2X2 + ... + ϵ\)
  • Data science interested in \(\hat{y}\), while in Empirical research / economics interested in \(β’\)s (e.g., the magnitude, statistical significance)

• Who will win the election?
• Is the climate changing?
• Why are women underrepresented in STEM degrees?
• What is the best way to prevent heart failure?
• What kind of topics are popular on Twitter?
• How many people visited in a given time period your website?
• …

• Description
  What happened?
• Explanation/Diagnostic
  Why did / does something happen?
• Prediction
  What is likely to happen in the future?
• Prescription
  What shall we do?

• Who will win the election? Prediction
• Is the climate changing? Description / Prediction
• Why are women underrepresented in STEM degrees? Explanation/Diagnostic
• What is the best way to prevent heart failure? Prescription
• What kind of topics are popular on Twitter? Description
• How many people visited in a given time period your website? Description
• …

• Exploratory
  • Following your gut (or other criteria) to some interesting results
  • Investigate datasets open-mindedly to uncover patterns, and understand data
  • Generate new insights
  • Create new hypotheses
• Confirmatory
  • Confirmatory data analysis tests specific hypotheses and confirms relationships with targeted analytical techniques
  • Validate or refute hypotheses
  • Analysis predefined

• Describing interesting patterns: use graphs, summaries, to understand subgroups, detect anomalies (“outliers”), understand the data

• Examples: boxplot, barplots, histograms, scatterplots…

• Confirmatory analysis

• You are a scientist and you want to determine if a new drug treatment is effective in reducing blood pressure compared to a placebo.

• Theory testing, hypothesis: new treatment better than placebo

• Analysis can be defined in advance: which outcome variables, how to sample from the population, which method?

• Inputs, but no outputs. Try to learn structure and relationships from these data, like detecting unobserved groups (clusters) from the data.
• A retail store has customer data such as age, income, number of purchases, and average purchase amount.
• By applying unsupervised learning techniques like clustering to the data, the retail store can gain valuable insights into customer segmentation, behavior patterns, and preferences.

Building a statistical model for predicting / estimating an output based on one or more inputs.

Most widely used machine learning methods are supervised

• Spam classifiers of e-mail
• Face recognizers over images
• Medical diagnosis systems for patients

Methods such as:

• Linear Regression -> week 4
• Logistic Regression -> week 5
• Decision trees / random forests -> week 7
• Support vector machines -> beyond the scope of this course
• Neural networks -> beyond the scope of this course
• and much more

• Some models are less flexible, they can produce a small range of shapes to estimate f
• Other models are more flexible and allow curve relationships

• Why would we ever choose to use a more restrictive method over a flexible approach?

• Measure of model accuracy: mean squared error:
  • \(MSE = \frac{1}{n} \sum^n_{i = 1}{(y_i - \hat{f}(x_i))^2}\)
    
  • \(MSE = \frac{1}{n} \sum^n_{i = 1}{(outcome_i - predicted_i)^2}\)
• Hence, MSE represents how wrong our prediction on average is.
• In machine / statistical learning, we obtain the MSE for a training set and a test set.

“The simple graph has brought more information to the data analyst’s mind than any other device.” — John Tukey

• Exploratory Data Visualization (week 2)
• Interactive visualization (week 6)

1. How wages vary over different groups (James et al. p. 2)

When thinking about important topics, such as whether to stay in school, it helps to know that more highly educated people tend to earn more, but also that there is no difference for top earners.

2. Healthcare during the Crimean War

Florence Nightingale’s Rose Diagram, created in 1858, is a an example of data visualization that played a significant role for improvements in healthcare during the Crimean War.

4. Telling Bronte sisters from Jane Austen

(Eder, M. et al. (2016). Stylometry with R: A Package for Computational Text Analysis. The R Journal 8:1)

4. Telling Bronte sisters from Jane Austen

Scholars fight over who wrote various songs (Wilhelmus), treatises (Caesar), plays (Shakespeare), etc., with shifting arguments. By counting words, we can sometimes identify the most likely author of a text, and we can explain exactly why we think that is the right answer.

• Data analysis and the accompanying visualisations yield insights and solves problems that could not be solved without them.
• On some level, humans do nothing but analyze data;
• They may not do it consistently, understandably, transparently, or correctly, however;
• aDAV help us process more data, and can keep us honest;
• aDAV can also exacerbate our biases when we are not careful;
• Data visualisations make people understand the data and remember the results.

Data analysis..

• …is usually part of a broader “pipeline”, may involve question formulation, budgeting, data collection, communication / dissemination, archiving, engineering, etc. – not covered here!
• …is often aided by visualisation;
• …can have different goals;
• …can have different modes;
• …can be useful, fun, but also dangerous;

In short, it makes sense to learn more about the background and ideas behind aDAV techniques!

• Lab 1 is on Thursday

• Let your teacher know if you can’t attend

• Submit the homework before 11:00am which will account for weeks’ one pass/fail grade

• Next lecture: next Tuesday about Data visualisation

Read the Course information and Preparation on the course website.