Beginning to 48:48 - Session 1; 48:49 to 1:36:25 - Session 2; 1:36:26 to 2:19:12 - Session 3
Session 1: Introduction to Multivariable Thinking - Nathan Tintle
Contexts and concepts discussed:
Cost of a college education - sources of variation;
Sand crab study - sources of variation
Session 2: Association and Confounding - Todd Swanson
Contexts and concepts discussed:
Do COVID vaccines work - relative risk, confounding variable, Simpson's paradox;
Seat belt usage - mosaic plots and adjusting for a third variable
Predicting a bird's body mass - association between two quantitative variables adjusted for a third variable
Session 3: Multivariable Thinking and Study Design with Social Justice Contexts - Laura Callis
Video on how teachers can set up Desmos for their own classes;
Password: tWY%GJ3d
Contexts and concepts discussed:
School to prison pipeline (three separate real world studies to investigate associations between student outcomes, teacher perceptions, and geography) - importance of study design and conclusions, proportions and percentages graphs and tables, descriptive statistics
Sources for studies:
Session 1: Multivariable Thinking in Designed Experiments - Karen McGaughey
Contexts and concepts discussed:
Tai Chi vs. stretching and the effect on functional reach - brainstorming sources of variation, importance of inclusion criteria, importance of random assignment, comparing two groups on a quantitative response, simulation-based inference (SBI);
Tai Chi vs. stretching vs. resistance training - comparing more than 2 groups on a quantitative response; SBI;
Storing strawberries - why to block, when to block, randomized complete block design (RCBD), accounting for block effects, SBI for a block design;
RCBD applet; data preloaded in applet
Session 2: Accounting for Other Variables in Designed Experiments - Soma Roy
Contexts and concepts discussed:
What predicts birth weights of babies? Brainstorming sources of variation
What predicts birth weights of baby skinks - two-variable designed experiment ( 2 x 2 factorial), completely randomized design (CRD), a One-Variable with 4 levels analysis vs. a Two-Variable additive model analysis vs. a Two-Variable Interaction model analysis, how a designed experiment partitions Total Sums of Squares and "adjusts" effects for other variables;
Session 3: Multivariable Thinking in Observational Studies - Beth Chance
Handout: Word/Google doc; PDF
Contexts and concepts discussed:
House prices in King County, WA from Kaggle – exploring scatterplots, transformations
MTL volume and sedentary behavior - adjusting for other variables in an observational study;
House prices in King County, WA - adjusting the relationship between price and number of bedrooms for different square footage; Simpson's paradox
KingCounty2 dataset (random sample of 2000)