Final Exam: Topics

Material covered

  • Lectures 16-21, Labs 7-8, AE 5-6

  • No formula sheet will be provided; all questions can be answered without one.

Coding

  • ggcorrplot(), cor(), pcor.test()

  • ggpairs()

  • How to subset columns/rows using brackets: i.e. iris[,1:4]

  • facet_wrap()

  • + geom_smooth(method = "lm")

  • tidy() and kable()

  • vif()

  • What does “tidy format” mean for a dataset?

  • glance() and augment()

  • What does the double colon specify? (e.g. dplyr::select())

  • lm(). How to fit a linear regression model?

  • glm(). How to fit a logistic regression model?

  • predict()

  • Differences in code for how to fit and predict outcomes for a linear vs. logistic regression model

Conceptual and Applied

  • Knowing common tests or models given the type of predictor and type of outcome (‘When to choose what method’ at the beginning of Lec 21).

  • Is correlation between X with Y the same as correlation between Y and X? What about the linear regression of Y on X versus the linear regression of X on Y?

  • Correlation coefficient \(\rho\): what type of relationship does it describe? Know the range and how to interpret positive/negative/zero values. Which statistic can we use to estimate \(\rho\)? (How to do this in R?)

    • Spearman vs. Pearson

  • Partial correlation. What is being tested?

  • What is the takeaway of Anscombe’s quartet?

  • Confounding definition

  • What is a regression line? What are residuals? What is error?

  • Know how to write out a linear model for a particular study, indexed by observation \(i = 1, \ldots, n\).Be able to define all terms in the model.

  • Interpreting coefficients from simple and multiple linear regression models. What’s the difference between simple vs. multiple linear regression?

  • Model assumptions of linear regression: be able to describe, and also how to check each one in R

  • In regression output, what are the individual coefficient statistics and p-values telling us? Which hypothesis test is being conducted for each p-value? Be able to write out the null and alternative hypotheses in symbols and in words.

    • For the global F-test, how would you interpret a p-value from the overall test? Be able to write out the null and alternative hypotheses in symbols and in words.

  • \(R^2\) and Adjusted \(R^2\): definitions and how to interpret. What’s the difference between them?

  • How to predict a new value based on estimated coefficients from a regression model? (e.g. writing out an expression)

  • How to construct a confidence interval for \(\hat\beta_k\)?

  • How does R treat categorical variables in a regression model? Know how to interpret a dummy variable from a regression model.

  • Interaction models: How to fit in R? How to interpret a main effect when an interaction effect is present in the model?

  • Multicollinearity: definition & how to detect (e.g. VIF)

  • How does logistic regression differ from linear regression?

  • In logistic regression, how do you interpret the estimated coefficients in terms of odds? How do you predict log-odds for a new observation given the estimated model coefficients?

  • How do you interpret the p-value for a single interaction term (like examples from Lecture 19)

  • How to predict a new value based on estimated coefficients from a logistic regression model? (e.g. writing out an expression) What does this value represent for this individual?