Clayton D., Hills M. Statistical Models in Epidemiology

Oxford: Oxford University Press, 2013. — 376 p.The most important concept in statistics is the probability model. Only by fully understanding this model can one fully understand statistical analysis. Utilizing models in epidemiology, the authors of this self-contained account have chosen to emphasize the role of likelihood. This approach to statistics is both simple and intuitively satisfying. More complex problems can be tackled by natural extensions of the simple methods. This exploration of the statistical basis of epidemiology has been written specifically for professionals and graduate students in epidemiology, clinical epidemiology, or biostatistics. The simple prerequisite--basic training in biology--assumes no previous knowledge and the mathematics is deliberately kept at a manageable level. Based on a highly successful course by two internationally known authors, this book explains the essentials of statistics for all epidemiologists.Probability Models and Likelihood
Probability models
Conditional probability models
Likelihood
Consecutive follow-up intervals
Rates
Time
Competing risks and selection
The Gaussian probability model
Approximate likelihoods
Likelihood, probability, and confidence
Null hypotheses and p-values
Small studies
Likelihoods for the rate ratio
Confounding and standardization
Comparison of rates within strata
Case-control studies
Likelihoods for the odds ratio
Comparison of odds within strata
Individually matched case-control studies
Tests for trend
The size of investigations
Introduction to regression models
Poisson and logistic regression
Testing hypotheses
Models for dose-response
More about interaction
Choice and interpretation of models
Additivity and synergism
Conditional logistic regression
Cox's regression analysis
Time-vary ing explanatory variables
Three examples
Nested case-control studies
Gaussian regression models
PostscriptAppendeces
Exponentials and logarithms
Some basic calculus
Approximate profile likelihoods
Table of the chi-squared distribution