Data Science in a(n Ever-Evolving) Box - MAA

Computation is foundational to all steps of the modern data analysis pipeline, from importing to tidying to transforming, visualizing, and modeling, to communicating data. However, this fact is generally not reflected in our traditional statistics curricula, which too often assume computing is something students can pick up on their own and courses should primarily cover the instruction of theory and applications. On the other hand, data science curricula put computation front and center. The rise of data science has given us an opportunity to re-examine our traditional curricula through a new, computational lens. In this talk, I will highlight how computation can support and enhance the teaching and learning of fundamental statistical concepts such as uncertainty quantification, and prediction. The talk will place these ideas within the context of a curriculum for an introductory data science and statistical thinking course that emphasizes explicit instruction in computing. Additionally, the talk will contextualize the course and its learning objectives in the larger context of an undergraduate statistics program.

Transcript

1. Data Science in a(n Evolving) Box Mine Çetinkaya-Rundel Duke University 2023-11-15
2. It All Starts with Math.
3. It All Starts with Math. The students don’t all start
4. Wickham, H., Çetinkaya-Rundel, M., & Grolemund, G. (2023). R for Data Science, 2nd Edition. Program Import Tidy Transform Visualize Model Communicate Understand Wickham, H., Çetinkaya-Rundel, M., & Grolemund, G. (2023). R for Data Science, 2nd Edition. MANY START WITH DATA SCIENCE
5. Communicate MANY START WITH DATA SCIENCE
6. Content Tooling Pedagogy
7. Content Tooling Pedagogy
8. Tooling Pedagogy Content in 3 examples
9. COURSE EVALUATIONS // INFERENCE 1 Image by Andreas Breitling from Pixabay
10. score rank ethnicity gender bty_avg <dbl> <chr> <chr> <chr> <dbl> 1 4.7 tenure track minority female 5 2 4.1 tenure track minority female 5 3 3.9 tenure track minority female 5 4 4.8 tenure track minority female 5 5 4.6 tenured not minority male 3 6 4.3 tenured not minority male 3 7 2.8 tenured not minority male 3 8 4.1 tenured not minority male 3.33 9 3.4 tenured not minority male 3.33 10 4.5 tenured not minority female 3.17 … … … … … … 463 4.1 tenure track minority female 5.33 evaluation score (1-5) beauty score (1-10) Hamermesh, Parker. “Beauty in the classroom: instructors pulchritude and putative pedagogical productivity”, Econ of Ed Review, Vol 24-4.
11. Estimate the difference in average evaluation scores of male and female faculty.
12. Approach 1: Using methods based on the Central Limit Theorem. t.test(score ~ gender, data = evals) Welch Two Sample t - test data: score by gender t = -2.7507, df = 398.7, p - value = 0.006218 alternative hypothesis: true difference in means between group female and group male is not equal to 0 95 percent conf i dence interval: -0.24264375 -0.04037194 sample estimates: mean in group female mean in group male 4.092821 4.234328
13. Approach 2: Using computational methods.
14. library(tidyverse) library(tidymodels) evals start with data Approach 2: Using computational methods. # A tibble: 463 × 23 course_id prof_id score rank ethnicity gender language age cls_perc_eval <int> <int> <dbl> <fct> <fct> <fct> <fct> <int> <dbl> 1 1 1 4.7 tenure… minority female english 36 55.8 2 2 1 4.1 tenure… minority female english 36 68.8 3 3 1 3.9 tenure… minority female english 36 60.8 4 4 1 4.8 tenure… minority female english 36 62.6 5 5 2 4.6 tenured not mino… male english 59 85 6 6 2 4.3 tenured not mino… male english 59 87.5 7 7 2 2.8 tenured not mino… male english 59 88.6 8 8 3 4.1 tenured not mino… male english 51 100 9 9 3 3.4 tenured not mino… male english 51 56.9 10 10 4 4.5 tenured not mino… female english 40 87.0 # ℹ 453 more rows # ℹ 14 more variables: cls_did_eval <int>, cls_students <int>, cls_level <fct>, # cls_profs <fct>, cls_credits <fct>, bty_f1lower <int>, bty_f1upper <int>, # bty_f2upper <int>, bty_m1lower <int>, bty_m1upper <int>, bty_m2upper <int>, # bty_avg <dbl>, pic_outf i t <fct>, pic_color <fct> # ℹ Use `print(n = ... )` to see more rows
15. library(tidyverse) library(tidymodels) evals |> specify(score ~ gender) Approach 2: Using computational methods. Response: score (numeric) Explanatory: gender (factor) # A tibble: 463 × 2 score gender <dbl> <fct> 1 4.7 female 2 4.1 female 3 3.9 female 4 4.8 female 5 4.6 male 6 4.3 male 7 2.8 male 8 4.1 male 9 3.4 male 10 4.5 female # ℹ 453 more rows # ℹ Use `print(n = ... )` to see more rows specify the model
16. library(tidyverse) library(tidymodels) evals |> specify(score ~ gender) |> generate(reps = 15000, type = "bootstrap") Approach 2: Using computational methods. Response: score (numeric) Explanatory: gender (factor) # A tibble: 6,945,000 × 3 # Groups: replicate [15,000] replicate score gender <int> <dbl> <fct> 1 1 4 female 2 1 3.1 male 3 1 5 male 4 1 4.4 male 5 1 3.5 female 6 1 4.5 female 7 1 4.5 male 8 1 4.9 male 9 1 4.4 male 10 1 3.5 male # ℹ 6,944,990 more rows # ℹ Use `print(n = ... )` to see more rows generate bootstrap samples
17. library(tidyverse) library(tidymodels) evals |> specify(score ~ gender) |> generate(reps = 15000, type = "bootstrap") |> calculate(stat = "diff in means", order = c("male", "female")) Approach 2: Using computational methods. Response: score (numeric) Explanatory: gender (factor) # A tibble: 15,000 × 2 replicate stat <int> <dbl> 1 1 0.230 2 2 0.134 3 3 0.100 4 4 0.230 5 5 0.128 6 6 0.201 7 7 0.168 8 8 0.130 9 9 -0.00490 10 10 0.123 # ℹ 14,990 more rows # ℹ Use `print(n = ... )` to see more rows calculate sample statistics
18. library(tidyverse) library(tidymodels) evals |> specify(score ~ gender) |> generate(reps = 15000, type = "bootstrap") |> calculate(stat = "diff in means", order = c("male", "female")) |> summarize(l = quantile(stat, 0.025), u = quantile(stat, 0.975)) Approach 2: Using computational methods. # A tibble: 1 × 2 l u <dbl> <dbl> 1 0.0431 0.242 summarize CI bounds
19. ✴ sampling variability ✴ inference via bootstrapping and randomization ✴ interpreting study results
20. FISHERIES OF THE WORLD // EXPLORATORY DATA ANALYSIS 2 Image by Nghĩa Đặng from Pixabay
22. ✴ data joins fisheries |> select(country) #> # A tibble: 82 × 1 #> country #> <chr> #> 1 Angola #> 2 Argentina #> 3 Australia #> 4 Bangladesh #> 5 Brazil #> 6 Cambodia #> 7 Cameroon #> 8 Canada #> 9 Chad #> 10 Chile # ℹ 72 more rows continents #> # A tibble: 245 × 2 #> country continent #> <chr> <chr> #> 1 Afghanistan Asia #> 2 Åland Islands Europe #> 3 Albania Europe #> 4 Algeria Africa #> 5 American Samoa Oceania #> 6 Andorra Europe #> 7 Angola Africa #> 8 Anguilla Americas #> 9 Antigua & Barbuda Americas #> 10 Argentina Americas #> # ℹ 235 more rows fisheries <- left_join(fisheries, continents) Joining with `by = join_by(country)`
23. ✴ data joins ✴ data science ethics fisheries |> filter(is.na(continent)) #> # A tibble: 3 × 5 #> country capture aquaculture total continent #> <chr> <dbl> <dbl> <dbl> <chr> #> 1 Democratic Republic of the Congo 237372 3161 240533 NA #> 2 Hong Kong 142775 4258 147033 NA #> 3 Myanmar 2072390 1017644 3090034 NA fisheries <- fisheries |> mutate( continent = case_when( country == "Democratic Republic of the Congo" ~ "Africa", country == "Hong Kong" ~ "Asia", country == "Myanmar" ~ "Asia", .default = continent ) )
24. ✴ data joins ✴ data science ethics ✴ critique ✴ improving data visualisations
25. ✴ data joins ✴ data science ethics ✴ critique ✴ improving data visualisations ✴ mapping
26. Project: Regional differences in average GPA and SAT Question: Exploring the regional differences in average GPA and SAT score across the US and the factors that could potentially explain them. Team: Mine’s Minions
27. COVID BRIEFINGS 2
29. ✴ web scraping ✴ text parsing ✴ data types ✴ regular expressions
30. ✴ web scraping ✴ text parsing ✴ data types ✴ regular expressions ✴ functions ✴ iteration
31. ✴ web scraping ✴ text parsing ✴ data types ✴ regular expressions ✴ functions ✴ iteration ✴ data visualisation ✴ interpretation
32. ✴ web scraping ✴ text parsing ✴ data types ✴ regular expressions ✴ functions ✴ iteration ✴ data visualisation ✴ interpretation ✴ text analysis
33. ✴ web scraping ✴ text parsing ✴ data types ✴ regular expressions ✴ functions ✴ iteration ✴ data visualisation ✴ interpretation ✴ text analysis ✴ data science ethics robotstxt::paths_allowed("https://www.gov.scot") #> www.gov.scot #> [1] TRUE
34. Project: Factors Most Important to University Ranking Question: Explore how various metrics (e.g., SAT/ACT scores, admission rate, region, Carnegie classification) predict rankings on the Niche College Ranking List. Team: 2cool4school
35. SPAM FILTERS // MODELING 3 Image by Gerd Altmann from Pixabay
36. ✴ logistic regression ✴ prediction
37. ✴ logistic regression ✴ prediction ✴ decision errors ✴ sensitivity / specificity ✴ intuition around loss functions
38. Project: Predicting League of Legends success Question: After 10 minutes into the game, whether a gold lead or an experienced lead was a better predictor of which team wins? Team: Blue Squirrels
39. Project: A Critique of Hollywood Relationship Stereotypes Question: How has the average age difference between two actors in an on-screen relationship changed over the years? Furthermore, do on-screen same-sex relationships have a different average age gap than on-screen heterosexual relationships? Team: team300
40. Content Tooling Pedagogy
41. creativity: assignments that make room for creativity peer feedback: at various stages of the project teams: weekly labs in teams + periodic team evaluations + term project in teams “minute paper”: weekly online quizzes ending with a brief reflection of the week’s material live coding: in every “lecture”, along with time for students to attempt exercises on their own
42. Çetinkaya-Rundel, Mine, Mine Dogucu, and Wendy Rummerfield. "The 5Ws and 1H of term projects in the introductory data science classroom." Statistics Education Research Journal 21.2 (2022): 4-4.
44. Content Pedagogy Tooling
45. + … Browser-based access to
46. ‣ Go to [URL to access RStudio in the browser] ‣ Start the project titled UN Votes
47. ‣ Go to [URL to access RStudio in the browser] ‣ Start the project titled UN Votes ‣ Open the Quarto document called unvotes.qmd
48. ‣ Go to [URL to access RStudio in the browser] ‣ Start the project titled UN Votes ‣ Open the Quarto document called unvotes.qmd ‣ Render the document and review the data visualization you just produced
49. ‣ Go to [URL to access RStudio in the browser] ‣ Start the project titled UN Votes ‣ Open the Quarto document called unvotes.qmd ‣ Render the document and review the data visualization you just produced ‣ Then, look for the character string “Turkey” in the code and replace it with another country of your choice ‣ Render again, and review how the voting patterns of the country you picked compare to the United States and the United Kingdom
50. Beckman, M. D., Çetinkaya-Rundel, M., Horton, N. J., Rundel, C. W., Sullivan, A. J., & Tackett, M. "Implementing version control with Git and GitHub as a learning objective in statistics and data science courses." Journal of Statistics and Data Science Education 29. (2021): S132-S144.
51. Content Tooling Pedagogy
52. Openness + Scalability
53. datasciencebox.org
54. datasciencebox.org
55. github.com/tidyverse/datascience-box
56. github.com/tidyverse/dsbox
57. AUDIENCE I have been teaching with R for a while, but I want to update my teaching materials I’m new to teaching with R and need to build up my course materials This teaching slide deck I came across is pretty cool, but I have no idea what type of course it belongs in
58. on COMMUNITY
59. sta199-f22-1.github.io EXAMPLE
60. Çetinkaya-Rundel, Mine, and Victoria Ellison. "A fresh look at introductory data science." Journal of Statistics and Data Science Education 29.sup1 (2021): S16-S26. SCHOLARSHIP
61. (N EVER-EVOLVING) mine-cetinkaya-rundel cetinkaya.mine@gmail.com fosstodon.org/@minecr 🔗 bit.ly/dsbox-evolving-maa thank you! minecr.bsky.social
62. Thank you for joining us for this MAA Distinguished Lecture Series presentation.