Statistics VSI

The technology of extracting meaning from data and of handling uncertainty. Mechanistic vs empirical models.

Main Page

Statistics

• Ubiquity of statistical ideas in society
• Technology not science: application of science
• A statistic vs discipline of statistics
• Lack of ambiguity in numerical data p24 “numbers have only one property”
• Average – dispersion – skewness – quantiles
• Incomplete data: why; incorrect data (poor definitions, misreading, device failing, nefarious actors)
• Observational vs experimental
• Randomization
• Experiments and experimental design
• Survey sampling
• Kolmogorov axioms; probability as degree of belief; subjective vs frequentist
• Bernoulli – binomial – Poisson – uniform – exponential – Gaussian
• Point estimation - maximum likelihood - least squares
• Prior – posterior – bias – error
• Bias – mse
• Interval estimate – credibility interval (Bayesian)
• Hypothesis testing – errors – significance testing (only null hypothesis)
• likelihood principal – repeated sampling principal – sufficiency principal
• Models
  • simple representation
  • Does it properly represent the underlying reality?
  • Construct models that are “good enough”
  • Mechanistic (theory) vs empirical (“convenient summaries of the important aspects of observed data”) models
  • Economics many mechanistic models based on theories
  • Prediction (smoking causes cancer) vs forecasting (likely rates of cancer)
  • Overfitting – generalizing
  • Linear – supervised – glm – clutering – trees – neural nets

Deets

• David Hand
• Volume 196
• Published 2008