Section 47 Regression Model: Next Level


47.1 Checkpoints in Statistical Modelling

  • Understand the system and background; physical or biological

  • Understand the objective and hypotheses

  • Translate the problem into statistical framework

  • Data collection procedures and designs - observation or experimental data

  • Characteristics of response and predictor variables - types of variables

  • Coding of the variables, units of variables

  • Presence and proportion of missing values

  • Exploratory Data Analysis

    • Summary Statistics of individual response and predictor variables
    • Plots to describe the variables
    • Relationship between the variables - Summary Statistics and Plots
    • Identify possible data entry errors or other inconsistencies
    • Identify the possible relationship and outcomes from possible model exploration

  • Statistical modelling


47.2 Advanced Topics in Regression Model

  • Validation, Cross-validation: Model selection based on training and testing data

  • Model selection when p >> n

  • Shrinkage (regularisation) method: Ridge, Lasso and Elastic Net

  • Dimension reduction method: Principal Component Regression (PCR), Partial Least Squares Regression (PLS)

  • Smoothing the relationship between x and y: Generalised Additive Model

  • Non-linear regression model: Growth curve

  • Regression model with Non-Gaussian error: Bernoulli, Binomial, Poisson, Negative Binomial

  • Regression model with pseudoreplication: Linear Mixed Model, Multilevel Model