An Introduction to Generalized Linear Models by Annette J. Dobson

By Annette J. Dobson

Carrying on with to stress numerical and graphical tools, An creation to Generalized Linear types, 3rd Edition presents a cohesive framework for statistical modeling. This re-creation of a bestseller has been up-to-date with Stata, R, and WinBUGS code in addition to 3 new chapters on Bayesian research.

Like its predecessor, this variation offers the theoretical historical past of generalized linear types (GLMs) earlier than targeting equipment for examining specific different types of information. It covers basic, Poisson, and binomial distributions; linear regression types; classical estimation and version becoming equipment; and frequentist equipment of statistical inference. After forming this beginning, the authors discover a number of linear regression, research of variance (ANOVA), logistic regression, log-linear versions, survival research, multilevel modeling, Bayesian types, and Markov chain Monte Carlo (MCMC) tools.

Using renowned statistical software program courses, this concise and available textual content illustrates functional methods to estimation, version becoming, and version comparisons. It comprises examples and workouts with entire information units for almost the entire versions lined.

Show description

Read Online or Download An Introduction to Generalized Linear Models PDF

Best probability & statistics books

Simulation and the Monte Carlo Method (Wiley Series in Probability and Statistics)

This booklet presents the 1st simultaneous insurance of the statistical elements of simulation and Monte Carlo tools, their commonalities and their transformations for the answer of a large spectrum of engineering and clinical difficulties. It comprises commonplace fabric often thought of in Monte Carlo simulation in addition to new fabric resembling variance relief concepts, regenerative simulation, and Monte Carlo optimization.

Confidence Intervals for Proportions and Related Measures of Effect Size

Self belief durations for Proportions and similar Measures of impression dimension illustrates using influence dimension measures and corresponding self assurance durations as extra informative choices to the main easy and prevalent value checks. The publication offers you a deep figuring out of what occurs whilst those statistical tools are utilized in events a long way faraway from the general Gaussian case.

Mathematical Methods of Statistics.

During this vintage of statistical mathematical concept, Harald Cramér joins the 2 significant traces of improvement within the box: whereas British and American statisticians have been constructing the technology of statistical inference, French and Russian probabilitists reworked the classical calculus of likelihood right into a rigorous and natural mathematical conception.

Additional info for An Introduction to Generalized Linear Models

Sample text

YN are independent. Here the link function is the identity function, g(µi ) = µi . This model is usually written in the form y = Xβ + e,   e1  ..  where e =  .  and the ei ’s are independent, identically distributed raneN dom variables with ei ∼ N(0, σ 2 ) for i = 1, . . , N . In this form, the linear component µ = Xβ represents the “signal” and e represents the “noise”, random variation or “error”.

4) the maximum likelihood estimates are θj = k yjk /Kj for j = 1 or 2. 0230. The maximum value of the log-likelihood function l1 will always be greater than or equal to that of l0 because one more parameter has been fitted. To decide whether the difference is statistically significant, we need to know the sampling distribution of the log-likelihood function. This is discussed in Chapter 4. If Y ∼ Po(θ) then E(Y ) = var(Y ) = θ. The estimate θ of E(Y ) is called the fitted value of Y . 4). Residuals form the basis of many methods for examining the adequacy of a model.

11) because {a(y)−E[a(Y )]}2 f (y; θ)dy = var[a(Y )] by definition. 9) gives var[a(Y )] = b′′ (θ)c′ (θ) − c′′ (θ)b′ (θ) . 4) and used to obtain the expected value and variance for other distributions in the exponential family. We also need expressions for the expected value and variance of the derivatives of the log-likelihood function. 3), the log-likelihood function for a distribution in the exponential family is l(θ; y) = a(y)b(θ) + c(θ) + d(y). The derivative of l(θ; y) with respect to θ is U (θ; y) = dl(θ; y) = a(y)b′ (θ) + c′ (θ).

Download PDF sample

Rated 4.09 of 5 – based on 39 votes