RMS Concepts to Master
- Assumptions of linear additive models
- Methods for checking these assumptions.
- Global vs. partial tests of association
- Multiple ways of computing test statistics in multiple regression models that were fitted using ordinary least squares
- Dummy variables and how their corresponding regression coefficients are interpreted
- Interpretation of interaction effects
- Assumptions of interaction tests
- Writing null hypotheses precisely in terms of parameters being tested
- Understanding tests for the overall association of a predictor with the response, and how to test sub-hypotheses such as linearity
- Combined partial tests for multiple predictors
- Combined tests for overall effects of a predictor when it interacts with other predictors
- Regression splines (linear, cubic, and restricted cubic) and knots
- How knots are chosen
- How the number of knots relates to the flexibility allowed for the fit
- Which if any terms of a predictor that is expanded into multiple constructed variables can be tested singly
- What tests of effects, interactions, and nonlinearity are powered to detect
- Nonparametric smoothers
- Problems with naive approaches of handing missing data
- The effect of changing how models are fitted based on looking at the data
- Deciding on the number of degrees of freedom to “spend” in a model, and where to spend them
- Understand regression to the mean
- Have an initial understanding of data reduction
- Elements of bootstrapping
- Model validation approaches and which methods of validation are most stringent.
- How to display a complex regression model to a non-statistician.
- How to make a complex nonlinear relationship a non-issue to the reader.
- A principle for estimating unknown parameters when least squares is not appropriate.
- What is a Wald statistic and a likelihood ratio statistic in general terms, and which one works better.
- When chi-square statistics are used instead of t or F statistics, and how to approximately relate a chi-square statistic to an F statistic.
- Exact interpretation of logistic model coefficients in the linear regression case.
- Assumptions of binary logistic regression.
- The value and use of a nonparametric smoother in examining logistic model assumptions or in determining shapes of relationships when Y is binary.
- How to convert between probabilities, odds, and log odds.
- Measures of predictive accuracy and predictive ability for binary logistic models.
- What is meant by an ordinal response variable and what is assumed about the data when you use a model or a rank test on an ordinal response.
- How to interpret coefficients in proportional odds models.
- What about odds ratios is assumed by the proportional odds model.
- How are ordinary nonparametric rank tests relate to the proportional odds model.
- What is the value of only using the ordering of Y.