Statistical Modeling and Data Visualization, Methods 2

Class schedules Spring, 2014

Class 1 (Jan 21): Course Introduction

Activities

  • quiz: wits and wagers
  • syllabus go-through, GitHub introduction
  • in-class computer discussion
  • class discussion: look at visualizations
  • Small group discussions
    • principles of effective data visualization
    • how to best collect WnW data
  • Big group discussion/wrap-up

Homework

  • Problem Set 1: Due Thursday, 2/13/2014 (by the beginning of class)
  • Create GitHub account, pull course repo into a directory on your machine. For instructions, see this video.
  • Read through the syllabus
  • Take CAOS test

Class 2 (Jan 23): Introduction to Regression

Activites

  • introduction to ggplot2 (15 min)
  • small groups: establish 5 specific criteria for creating good data visualizations (10 min)
  • big group: consensus criteria (15 min)
  • mini-lecture: introduction to regression (30 min)

Homework

  • Problem Set 1: Due Thursday, 2/13/2014 (by the beginning of class)
  • Create a short reproducible document (using knitr) that describes the basic structure of a dataset and summarizes some key features of the data using a few key tables and figures. Choose a dataset from these datasets or the ones in the class Google Drive. If your dataset has a lot of variables, focus on a subset of them – less than 6 or so – for the purposes of this exercise. Your write-up should answer the following questions:
    • What is the background/context for this data?
    • How many observations are there?
    • What is the unit of observation?
    • Is there any missing data? If so, are there patterns to the missingness?
    • What are the key variables and what do their distributions look like?
    • Is there a pair of variables that might work well for a Simple Linear Regression? (You don’t necessarily need to run one, but you could.)
    • Are there any obvious outliers in the data?

Class 3 (Jan 28): Geometry of regression and least squares

Activities

  • homework discussion/questions (10 min)
  • warm-ups (10 min)
  • mini-lecture: least squares and geometry of regression (30 min)
  • lab: OpenIntro Lab 7. (20 min)
  • exercise: Everyone make a guess at minimal RSS after three tries with the plot_ss(). Take averages in groups. Compare to minimal RSS based on summary(lm()) output.

Homework

  • Problem Set 1: Due Thursday, 2/13/2014 (by the beginning of class)
  • Add one or two simple linear regressions to your dataset write-up.
  • Finish OpenIntro Lab 7.
  • Install the HSAUR2 package, read up on and explore the BtheB dataset (hint: ?BtheB). Be prepared to describe the dataset and answer questions about it in next class.

Class 4 (Jan 30): Hands-on SLR practice

Activities

  • warm-ups (10 min)
  • introductions (5 min)
  • OpenIntro lab 7 questions (5 min)
  • More dataset descriptions/results, including BtheB (5 min)
  • small groups: Formulate and fit a reasonable SLR model to BtheB dataset. (30 min)
  • whole class: Present regressions. Talk about different model formulations and results. (15 min)

Homework

  • Problem Set 1: Due Thursday, 2/13/2014 (by the beginning of class)
  • In your small groups, create a write-up for a simple analysis of the BtheB dataset. Each group should create a repository on GitHub for their analysis. The .Rnw or .Rmd file that you use should be in that repository and every member of the group should have at least one commit or push to the repository before the next class. (You should not commit any additional files, like the .aux files from LaTeX compiling, just the files that are needed to comile your analysis.) Here is a minimal list of things that should be included in your write-up. I encourage you to push beyond just this list, however.
    • A few sentences of background/context for the BtheB dataset.
    • A quantitative and/or visual description of what variables you chose to use for your analysis, along with a hypothesis (or two) that you will be testing.
    • A description of the characteristics of the missing data (including a figure if needed) and a statement and justification as to whether your group is concerned about the missingness having an impact on your analysis.
    • Results, with interpretation, of output from an SLR model. We haven’t discussed yet using binary predictors or X variables, but feel free to include them. The interpretation is very similar. “For a one unit change in X, …”
  • Read this description of how to write and use functions in R.

Class 5 (Feb 4): R^2, ANOVA

Activities

  • warm-ups (10 min)
  • lecture: SLR final concepts (40 min)
  • writing functions in R (15 min)

Homework

  • Problem Set 1: Due Thursday, 2/13/2014 (by the beginning of class)
  • Create a “slr()” R function that takes x and y vectors and outputs a list with two objects: (1) a fitted lm() object and (2) by-hand betas (calculated by likelihood or formulae). Try to write this as a function, but if you have trouble, then just write it as a few lines of R code and create an object as described.
  • Use this new slr() function/code to refit the SLR models in your dataset writeup. Compare the results and make sure they are returning the same thing.

Class 6 (Feb 6): Version control

Activities

  • warm ups (15 min)
  • demo: GitHub and RStudio demo (10 min)
  • small groups: compare slr() functions from homework. summarize similarities and differences in approaches. Was there a consensus “best” approach? (20 min)

Homework

  • Problem Set 1: Due Thursday, 2/13/2014 (by the beginning of class)

Class 7 (Feb 11): MLR interpretations

Activities

  • warm ups: wits and wagers (15 min)
  • small groups: collecting data on wits and wagers (10 min)
  • homework update
  • MLR coefficient interpretation and matrix notation (20 min)
  • MLR example walk-through (15 min)

Homework

  • Problem Set 1: Due Tuesday, 2/25/2014 (by the beginning of class)
  • Add a fitted MLR to your dataset write-up. State the model, in equation form. Describe it in words. Interpret your fitted coefficients.

Class 8 (Feb 13):

SNOW DAY, CLASS CANCELLED.

Class 9 (Feb 20): MLR estimation and notation

Activities

  • lecture: matrix formulation of MLR
  • small groups: visualization crititque

Homework

  • Problem Set 1: Due Tuesday, 2/25/2014 (by the beginning of class)
  •  

Class 10 (Feb 25): MLR collinearity

Activities

  • wits and wagers CI exercise
  • lecture: hat matrix, collinearity

Homework

  • Problem Set 1: Due TODAY at beginning of class

Class 11 (Feb 27): MLR categorical variables

Activities

  • Mid-semester course evaluations (25 min)
  • lecture: MLR Categorical variables

Homework

Class 12 (Mar 4): MLR inference and testing

Activities

  • warm-ups
  • lecture: MLR inference and testing (45 min)

Homework

  • Problem Set 2, Due Friday 3/7, 5pm.
  • Include one global F test (with written interpretation) in your problem set 2 write-up.

Class 13 (Mar 6): MLR GLobal F Tests

Activities

Homework

  • Problem Set 2, Due Friday 3/7, 5pm.
  • Visualization assignment: Please dig through the scientific literature (no limit on the discipline, but I’d encourage you to look in the literature for your field) and find a paper that uses some form of multiple linear regression models. The paper should be about modeling a continuous outcome (no logistic/log-linear/etc.. models) and could incorporate some more fancy modeling things than we’ve talked about (e.g. spline terms, polynomials, random effects). The paper should have a figure/graphic in it that attempts to visualize features of the regression. Post a link to the paper on Piazza (either a URL or a PDF file), with a note about which graphic you think is most relevant to the discussion, and describe what tool(s) you used to find the paper (e.g. Google Scholar, PubMed, Scopus, …). Feel free to comment/start discussion on papers that other students post, but everyone is expected to find their own paper. I’ll choose one that we’ll talk about in more detail on Tuesday.

Class 14 (Mar 11): MLR non-parametric inference

Activities

  • warm-ups (visualization critique)
  • lecture: simulation and inference from resampling
  • permutation test lab ( html source )

Homework

  • Start a new data analysis write-up, include a brief summary from the first round.
  • Add bootstrapped inference to the new write-up.

Class 15 (Mar 15): MLR diagnostics

Activities

  • Data Fest promo with Andrew
  • wits and wagers
  • review of course expectations
  • demo and lecture: regression diagnostics

Homework

  • Dataset for final project proposed by Wednesday, March 26

Class 16 (Mar 25): MLR diagnostics (continued)

Activities

  • wits and wagers
  • lecture: regression diagnostics
  • project and class schedule check-in
  • in groups: dataset discussions

Homework

  • Dataset for final project proposed by Wednesday, March 26
  • Individual project topics proposed by Wednesday, April 2
  • Problem set 3 due Friday, April 4 at 5pm.

Class 17 (Mar 27): MLR model selection

Activities

  • lecture: model selection
  • FEV dataset walk-through analysis in class
  • group work on projects

Homework

  • Individual project topics proposed by Wednesday, April 2
  • Problem set 3 due Friday, April 4 at 5pm.

Class 18 (April 1): MLR interactions, variable transformations

Activities

  • lecture: interaction and variable transformations
  • group work on projects

Homework

  • Individual project topics proposed by Wednesday, April 2
  • Problem set 3 due Friday, April 4 at 5pm.

Class 19 (April 3): Spline models

Activities

  • lecture: finish interaction and variable transformations
  • quiz
  • lecture: spline models

Homework

  • Problem set 3 due Friday, April 4 at 5pm.

Class 20 (April 8): GLMs and logistic regression

Activities

  • project update: comments on group drafts and data summaries, timeline review, Google Drive system
  • lecture: GLM and logistic regression
  • group work on projects

Homework

  • Mon Apr 14: Draft of individual data analysis due (hand in PDF in your group’s project folder on Google Drive)

Class 21 (April 10): Logistic regression competition

Activities

  • logistic regression competition

Homework

  • Mon Apr 14: Draft of individual data analysis due (hand in PDF in your group’s project folder on Google Drive)