DOCSLIB.ORG

Generalized Linear Models the General Linear Model Approach Used So Far

Total Page：16

File Type：pdf, Size：1020Kb

Download full-text PDF Read full-text
Generalized Linear Models the General Linear Model Approach Used So Far Generalized Linear Models The general linear model approach used so far: • consists of specifications for the mean and covariance struc- ture; • mean response is a linear function of parameters and covari- ates; • has strong theoretical support when the responses are mul- tivariate normal; • works well when the response is continuous. 1 Generalized Linear Models When the response is discrete, or the mean response is naturally modeled as nonlinear in the parameters, we need an alternative. Other members of the exponential family of distributions provide various alternatives. Each distribution has a natural way to relate the mean response to the parameters and covariates, and implies a relationship be- tween the mean response and the variance. Begin with independent responses; longitudinal data handled by allowing for dependence or with random effects. 2 Distributions The most important distributions are: • normal • Bernoulli • binomial • Poisson where the last three are used for modeling count data. 3 Variance Functions For each of these, the variance is in the form φv(µ) for some dispersion parameter φ and variance function v(µ): • normal: Var(Y ) = σ2, so v(µ) = 1, φ = σ2; • Bernoulli: E(Y ) = p, Var(Y ) = p(1 − p), so v(µ) = µ(1 − µ), φ = 1; • binomial: E(Y ) = np, Var(Y ) = np(1−p), so v(µ) = µ(n − µ)/n, φ = 1; • Poisson: E(Y ) = Var(Y ) = µ, so v(µ) = µ, φ = 1. 4 Link Functions The link function links the mean response to the parameters and covariates: 0 g(µ) = η = β1X1 + β2X2 + ··· + βpXp = X β. Each distribution has a canonical link function, derived from its natural parametrization in the exponential family. Other link functions can also be used, when the subject matter suggests one. Distribution Variance Function, v(µ) Canonical Link Normal v(µ) = 1 Identity: µ = η µ Bernoulli v(µ) = µ(1 − µ) Logit: log 1−µ = η Poission v(µ) = µ Log: log(µ) = η 5 Logistic Regression • Response Y is Bernoulli: P(Y = 1) = 1 − P(Y = 0) = µ. • Variance function: Var(Y ) = µ(1 − µ) (whence φ = 1). • Covariates: µ = µ(X, β). • Require 0 ≤ µ(X, β) ≤ 1: rules out linear model µ = X0β. 6 Logistic Link Function The log odds ratio µ ! log = logit(µ) = η 1 − µ can be solved for µ: exp(η) µ = . 1 + exp(η) Any value for the log odds gives µ satisfying 0 < µ < 1. So a linear model for logit(µ) η = X0β always gives valid values of µ. 7 Interpretation of Parameters If one covariate changes by +1 and the others are not changed, η changes by the corresponding β coefficient. If µ ≈ 0, logit(µ) ≈ log(µ), so log(µ) also changes by approxi- mately β. If β is also small, a change of β in log(µ) is approximately a fractional change of β in µ. For example, if β = 0.1, a change of +1 in the covariate implies a 10% increase in µ. At the other extreme, if µ ≈ 1, logit(µ) ≈ − log(1 − µ). So now, if β is still 0.1, a change of +1 in the covariate implies a 10% decrease in 1 − µ. 8 Binomial Case If Y1,Y2,...,Ym are independent Bernoulli with the same covari- ates (or at least, the same µ), then Y = Y1 + Y2 + ··· + Ym is Bin(m, µ). • The logistic link function is still reasonable. • The binomial variance function is Var(Y ) = mµ(1 − µ). 9 • In practice, Y is often over-dispersed: Var(Y ) > mµ(1 − µ). • The variance function therefore often includes the dispersion parameter: v(µ) = φmµ(1 − µ). • Over-dispersion can be caused by: dependence: conditionally on the covariates, the individual Yjs are not independent (e.g., common cause of failure); heterogeneity: the individual Yjs do not have exactly the same covariates, or covariates are measured with error. 10.
Load more

Recommended publications
  • Generalized Linear Models and Generalized Additive Models
    00:34 Friday 27th February, 2015 Copyright ©Cosma Rohilla Shalizi; do not distribute without permission updates at http://www.stat.cmu.edu/~cshalizi/ADAfaEPoV/ Chapter 13 Generalized Linear Models and Generalized Additive Models [[TODO: Merge GLM/GAM and Logistic Regression chap- 13.1 Generalized Linear Models and Iterative Least Squares ters]] [[ATTN: Keep as separate Logistic regression is a particular instance of a broader kind of model, called a gener- chapter, or merge with logis- alized linear model (GLM). You are familiar, of course, from your regression class tic?]] with the idea of transforming the response variable, what we’ve been calling Y , and then predicting the transformed variable from X . This was not what we did in logis- tic regression. Rather, we transformed the conditional expected value, and made that a linear function of X . This seems odd, because it is odd, but it turns out to be useful. Let’s be specific. Our usual focus in regression modeling has been the condi- tional expectation function, r (x)=E[Y X = x]. In plain linear regression, we try | to approximate r (x) by β0 + x β. In logistic regression, r (x)=E[Y X = x] = · | Pr(Y = 1 X = x), and it is a transformation of r (x) which is linear. The usual nota- tion says | ⌘(x)=β0 + x β (13.1) · r (x) ⌘(x)=log (13.2) 1 r (x) − = g(r (x)) (13.3) defining the logistic link function by g(m)=log m/(1 m). The function ⌘(x) is called the linear predictor. − Now, the first impulse for estimating this model would be to apply the transfor- mation g to the response.
    [Show full text]
  • Stochastic Process - Introduction
    Stochastic Process - Introduction • Stochastic processes are processes that proceed randomly in time. • Rather than consider fixed random variables X, Y, etc. or even sequences of i.i.d random variables, we consider sequences X0, X1, X2, …. Where Xt represent some random quantity at time t. • In general, the value Xt might depend on the quantity Xt-1 at time t-1, or even the value Xs for other times s < t. • Example: simple random walk . week 2 1 Stochastic Process - Definition • A stochastic process is a family of time indexed random variables Xt where t belongs to an index set. Formal notation, { t : ∈ ItX } where I is an index set that is a subset of R. • Examples of index sets: 1) I = (-∞, ∞) or I = [0, ∞]. In this case Xt is a continuous time stochastic process. 2) I = {0, ±1, ±2, ….} or I = {0, 1, 2, …}. In this case Xt is a discrete time stochastic process. • We use uppercase letter {Xt } to describe the process. A time series, {xt } is a realization or sample function from a certain process. • We use information from a time series to estimate parameters and properties of process {Xt }. week 2 2 Probability Distribution of a Process • For any stochastic process with index set I, its probability distribution function is uniquely determined by its finite dimensional distributions. •The k dimensional distribution function of a process is defined by FXX x,..., ( 1 x,..., k ) = P( Xt ≤ 1 ,..., xt≤ X k) x t1 tk 1 k for anyt 1 ,..., t k ∈ I and any real numbers x1, …, xk .
    [Show full text]
  • Variance Function Regressions for Studying Inequality
    Variance Function Regressions for Studying Inequality The Harvard community has made this article openly available. Please share how this access benefits you. Your story matters Citation Western, Bruce and Deirdre Bloome. 2009. Variance function regressions for studying inequality. Working paper, Department of Sociology, Harvard University. Citable link http://nrs.harvard.edu/urn-3:HUL.InstRepos:2645469 Terms of Use This article was downloaded from Harvard University’s DASH repository, and is made available under the terms and conditions applicable to Open Access Policy Articles, as set forth at http:// nrs.harvard.edu/urn-3:HUL.InstRepos:dash.current.terms-of- use#OAP Variance Function Regressions for Studying Inequality Bruce Western1 Deirdre Bloome Harvard University January 2009 1Department of Sociology, 33 Kirkland Street, Cambridge MA 02138. E-mail: [email protected]. This research was supported by a grant from the Russell Sage Foundation. Abstract Regression-based studies of inequality model only between-group differ- ences, yet often these differences are far exceeded by residual inequality. Residual inequality is usually attributed to measurement error or the in- fluence of unobserved characteristics. We present a regression that in- cludes covariates for both the mean and variance of a dependent variable. In this model, the residual variance is treated as a target for analysis. In analyses of inequality, the residual variance might be interpreted as mea- suring risk or insecurity. Variance function regressions are illustrated in an analysis of panel data on earnings among released prisoners in the Na- tional Longitudinal Survey of Youth. We extend the model to a decomposi- tion analysis, relating the change in inequality to compositional changes in the population and changes in coefficients for the mean and variance.
    [Show full text]
  • Flexible Signal Denoising Via Flexible Empirical Bayes Shrinkage
    Journal of Machine Learning Research 22 (2021) 1-28 Submitted 1/19; Revised 9/20; Published 5/21 Flexible Signal Denoising via Flexible Empirical Bayes Shrinkage Zhengrong Xing [email protected] Department of Statistics University of Chicago Chicago, IL 60637, USA Peter Carbonetto [email protected] Research Computing Center and Department of Human Genetics University of Chicago Chicago, IL 60637, USA Matthew Stephens [email protected] Department of Statistics and Department of Human Genetics University of Chicago Chicago, IL 60637, USA Editor: Edo Airoldi Abstract Signal denoising—also known as non-parametric regression—is often performed through shrinkage estima- tion in a transformed (e.g., wavelet) domain; shrinkage in the transformed domain corresponds to smoothing in the original domain. A key question in such applications is how much to shrink, or, equivalently, how much to smooth. Empirical Bayes shrinkage methods provide an attractive solution to this problem; they use the data to estimate a distribution of underlying “effects,” hence automatically select an appropriate amount of shrinkage. However, most existing implementations of empirical Bayes shrinkage are less flexible than they could be—both in their assumptions on the underlying distribution of effects, and in their ability to han- dle heteroskedasticity—which limits their signal denoising applications. Here we address this by adopting a particularly flexible, stable and computationally convenient empirical Bayes shrinkage method and apply- ing it to several signal denoising problems. These applications include smoothing of Poisson data and het- eroskedastic Gaussian data. We show through empirical comparisons that the results are competitive with other methods, including both simple thresholding rules and purpose-built empirical Bayes procedures.
    [Show full text]
  • Generalized Linear Models
    Generalized Linear Models A generalized linear model (GLM) consists of three parts. i) The first part is a random variable giving the conditional distribution of a response Yi given the values of a set of covariates Xij. In the original work on GLM’sby Nelder and Wedderburn (1972) this random variable was a member of an exponential family, but later work has extended beyond this class of random variables. ii) The second part is a linear predictor, i = + 1Xi1 + 2Xi2 + + ··· kXik . iii) The third part is a smooth and invertible link function g(.) which transforms the expected value of the response variable, i = E(Yi) , and is equal to the linear predictor: g(i) = i = + 1Xi1 + 2Xi2 + + kXik. ··· As shown in Tables 15.1 and 15.2, both the general linear model that we have studied extensively and the logistic regression model from Chapter 14 are special cases of this model. One property of members of the exponential family of distributions is that the conditional variance of the response is a function of its mean, (), and possibly a dispersion parameter . The expressions for the variance functions for common members of the exponential family are shown in Table 15.2. Also, for each distribution there is a so-called canonical link function, which simplifies some of the GLM calculations, which is also shown in Table 15.2. Estimation and Testing for GLMs Parameter estimation in GLMs is conducted by the method of maximum likelihood. As with logistic regression models from the last chapter, the generalization of the residual sums of squares from the general linear model is the residual deviance, Dm 2(log Ls log Lm), where Lm is the maximized likelihood for the model of interest, and Ls is the maximized likelihood for a saturated model, which has one parameter per observation and fits the data as well as possible.
    [Show full text]
  • Variance Function Estimation in Multivariate Nonparametric Regression by T
    Variance Function Estimation in Multivariate Nonparametric Regression by T. Tony Cai, Lie Wang University of Pennsylvania Michael Levine Purdue University Technical Report #06-09 Department of Statistics Purdue University West Lafayette, IN USA October 2006 Variance Function Estimation in Multivariate Nonparametric Regression T. Tony Cail, Michael Levine* Lie Wangl October 3, 2006 Abstract Variance function estimation in multivariate nonparametric regression is consid- ered and the minimax rate of convergence is established. Our work uses the approach that generalizes the one used in Munk et al (2005) for the constant variance case. As is the case when the number of dimensions d = 1, and very much contrary to the common practice, it is often not desirable to base the estimator of the variance func- tion on the residuals from an optimal estimator of the mean. Instead it is desirable to use estimators of the mean with minimal bias. Another important conclusion is that the first order difference-based estimator that achieves minimax rate of convergence in one-dimensional case does not do the same in the high dimensional case. Instead, the optimal order of differences depends on the number of dimensions. Keywords: Minimax estimation, nonparametric regression, variance estimation. AMS 2000 Subject Classification: Primary: 62G08, 62G20. Department of Statistics, The Wharton School, University of Pennsylvania, Philadelphia, PA 19104. The research of Tony Cai was supported in part by NSF Grant DMS-0306576. 'Corresponding author. Address: 250 N. University Street, Purdue University, West Lafayette, IN 47907. Email: [email protected]. Phone: 765-496-7571. Fax: 765-494-0558 1 Introduction We consider the multivariate nonparametric regression problem where yi E R, xi E S = [0, Ild C Rd while a, are iid random variables with zero mean and unit variance and have bounded absolute fourth moments: E lail 5 p4 < m.
    [Show full text]
  • Variance Function Program
    Variance Function Program Version 15.0 (for Windows XP and later) July 2018 W.A. Sadler 71B Middleton Road Christchurch 8041 New Zealand Ph: +64 3 343 3808 e-mail: [email protected] (formerly at Nuclear Medicine Department, Christchurch Hospital) Contents Page Variance Function Program 15.0 1 Windows Vista through Windows 10 Issues 1 Program Help 1 Gestures 1 Program Structure 2 Data Type 3 Automation 3 Program Data Area 3 User Preferences File 4 Auto-processing File 4 Graph Templates 4 Decimal Separator 4 Screen Settings 4 Scrolling through Summaries and Displays 4 The Variance Function 5 Variance Function Output Examples 8 Variance Stabilisation 11 Histogram, Error and Biological Variation Plots 12 Regression Analysis 13 Limit of Blank (LoB) and Limit of Detection (LoD) 14 Bland-Altman Analysis 14 Appendix A (Program Delivery) 16 Appendix B (Program Installation) 16 Appendix C (Program Removal) 16 Appendix D (Example Data, SD and Regression Files) 16 Appendix E (Auto-processing Example Files) 17 Appendix F (History) 17 Appendix G (Changes: Version 14.0 to Version 15.0) 18 Appendix H (Version 14.0 Bug Fixes) 19 References 20 Variance Function Program 15.0 1 Variance Function Program 15.0 The variance function (the relationship between variance and the mean) has several applications in statistical analysis of medical laboratory data (eg. Ref. 1), but the single most important use is probably the construction of (im)precision profile plots (2). This program (VFP.exe) was initially aimed at immunoassay data which exhibit relatively extreme variance relationships. It was based around the “standard” 3-parameter variance function, 2 J σ = (β1 + β2U) 2 where σ denotes variance, U denotes the mean, and β1, β2 and J are the parameters (3, 4).
    [Show full text]
  • Robust Bayesian General Linear Models ⁎ W.D
    www.elsevier.com/locate/ynimg NeuroImage 36 (2007) 661–671 Robust Bayesian general linear models ⁎ W.D. Penny, J. Kilner, and F. Blankenburg Wellcome Department of Imaging Neuroscience, University College London, 12 Queen Square, London WC1N 3BG, UK Received 22 September 2006; revised 20 November 2006; accepted 25 January 2007 Available online 7 May 2007 We describe a Bayesian learning algorithm for Robust General Linear them from the data (Jung et al., 1999). This is, however, a non- Models (RGLMs). The noise is modeled as a Mixture of Gaussians automatic process and will typically require user intervention to rather than the usual single Gaussian. This allows different data points disambiguate the discovered components. In fMRI, autoregressive to be associated with different noise levels and effectively provides a (AR) modeling can be used to downweight the impact of periodic robust estimation of regression coefficients. A variational inference respiratory or cardiac noise sources (Penny et al., 2003). More framework is used to prevent overfitting and provides a model order recently, a number of approaches based on robust regression have selection criterion for noise model order. This allows the RGLM to default to the usual GLM when robustness is not required. The method been applied to imaging data (Wager et al., 2005; Diedrichsen and is compared to other robust regression methods and applied to Shadmehr, 2005). These approaches relax the assumption under- synthetic data and fMRI. lying ordinary regression that the errors be normally (Wager et al., © 2007 Elsevier Inc. All rights reserved. 2005) or identically (Diedrichsen and Shadmehr, 2005) distributed. In Wager et al.
    [Show full text]
  • Generalized Linear Models Outline for Today
    Generalized linear models Outline for today • What is a generalized linear model • Linear predictors and link functions • Example: estimate a proportion • Analysis of deviance • Example: fit dose-response data using logistic regression • Example: fit count data using a log-linear model • Advantages and assumptions of glm • Quasi-likelihood models when there is excessive variance Review: what is a (general) linear model A model of the following form: Y = β0 + β1X1 + β2 X 2 + ...+ error • Y is the response variable • The X ‘s are the explanatory variables • The β ‘s are the parameters of the linear equation • The errors are normally distributed with equal variance at all values of the X variables. • Uses least squares to fit model to data, estimate parameters • lm in R The predicted Y, symbolized here by µ, is modeled as µ = β0 + β1X1 + β2 X 2 + ... What is a generalized linear model A model whose predicted values are of the form g(µ) = β0 + β1X1 + β2 X 2 + ... • The model still include a linear predictor (to right of “=”) • g(µ) is the “link function” • Wide diversity of link functions accommodated • Non-normal distributions of errors OK (specified by link function) • Unequal error variances OK (specified by link function) • Uses maximum likelihood to estimate parameters • Uses log-likelihood ratio tests to test parameters • glm in R The two most common link functions Log • used for count data η η = logµ The inverse function is µ = e Logistic or logit • used for binary data µ eη η = log µ = 1− µ The inverse function is 1+ eη In all cases log refers to natural logarithm (base e).
    [Show full text]
  • Examining Residuals
    Stat 565 Examining Residuals Jan 14 2016 Charlotte Wickham stat565.cwick.co.nz So far... xt = mt + st + zt Variable Trend Seasonality Noise measured at time t Estimate and{ subtract off Should be left with this, stationary but probably has serial correlation Residuals in Corvallis temperature series Temp - loess smooth on day of year - loess smooth on date Your turn Now I have residuals, how could I check the variance doesn't change through time (i.e. is stationary)? Is the variance stationary? Same checks as for mean except using squared residuals or absolute value of residuals. Why? var(x) = 1/n ∑ ( x - μ)2 Converts a visual comparison of spread to a visual comparison of mean. Plot squared residuals against time qplot(date, residual^2, data = corv) + geom_smooth(method = "loess") Plot squared residuals against season qplot(yday, residual^2, data = corv) + geom_smooth(method = "loess", size = 1) Fitted values against residuals Looking for mean-variance relationship qplot(temp - residual, residual^2, data = corv) + geom_smooth(method = "loess", size = 1) Non-stationary variance Just like the mean you can attempt to remove the non-stationarity in variance. However, to remove non-stationary variance you divide by an estimate of the standard deviation. Your turn For the temperature series, serial dependence (a.k.a autocorrelation) means that today's residual is dependent on yesterday's residual. Any ideas of how we could check that? Is there autocorrelation in the residuals? > corv$residual_lag1 <- c(NA, corv$residual[-nrow(corv)]) > head(corv) . residual residual_lag1 . 1.5856663 NA xt-1 = lag 1 of xt . -0.4928295 1.5856663 .
    [Show full text]
  • Estimating Variance Functions for Weighted Linear Regression
    Kansas State University Libraries New Prairie Press Conference on Applied Statistics in Agriculture 1992 - 4th Annual Conference Proceedings ESTIMATING VARIANCE FUNCTIONS FOR WEIGHTED LINEAR REGRESSION Michael S. Williams Hans T. Schreuder Timothy G. Gregoire William A. Bechtold See next page for additional authors Follow this and additional works at: https://newprairiepress.org/agstatconference Part of the Agriculture Commons, and the Applied Statistics Commons This work is licensed under a Creative Commons Attribution-Noncommercial-No Derivative Works 4.0 License. Recommended Citation Williams, Michael S.; Schreuder, Hans T.; Gregoire, Timothy G.; and Bechtold, William A. (1992). "ESTIMATING VARIANCE FUNCTIONS FOR WEIGHTED LINEAR REGRESSION," Conference on Applied Statistics in Agriculture. https://doi.org/10.4148/2475-7772.1401 This is brought to you for free and open access by the Conferences at New Prairie Press. It has been accepted for inclusion in Conference on Applied Statistics in Agriculture by an authorized administrator of New Prairie Press. For more information, please contact [email protected]. Author Information Michael S. Williams, Hans T. Schreuder, Timothy G. Gregoire, and William A. Bechtold This is available at New Prairie Press: https://newprairiepress.org/agstatconference/1992/proceedings/13 Conference on153 Applied Statistics in Agriculture Kansas State University ESTIMATING VARIANCE FUNCTIONS FOR WEIGHTED LINEAR REGRESSION Michael S. Williams Hans T. Schreuder Multiresource Inventory Techniques Rocky Mountain Forest and Range Experiment Station USDA Forest Service Fort Collins, Colorado 80526-2098, U.S.A. Timothy G. Gregoire Department of Forestry Virginia Polytechnic Institute and State University Blacksburg, Virginia 24061-0324, U.S.A. William A. Bechtold Research Forester Southeastern Forest Experiment Station Forest Inventory and Analysis Project USDA Forest Service Asheville, North Carolina 28802-2680, U.S.A.
    [Show full text]
  • Stat 714 Linear Statistical Models
    STAT 714 LINEAR STATISTICAL MODELS Fall, 2010 Lecture Notes Joshua M. Tebbs Department of Statistics The University of South Carolina TABLE OF CONTENTS STAT 714, J. TEBBS Contents 1 Examples of the General Linear Model 1 2 The Linear Least Squares Problem 13 2.1 Least squares estimation . 13 2.2 Geometric considerations . 15 2.3 Reparameterization . 25 3 Estimability and Least Squares Estimators 28 3.1 Introduction . 28 3.2 Estimability . 28 3.2.1 One-way ANOVA . 35 3.2.2 Two-way crossed ANOVA with no interaction . 37 3.2.3 Two-way crossed ANOVA with interaction . 39 3.3 Reparameterization . 40 3.4 Forcing least squares solutions using linear constraints . 46 4 The Gauss-Markov Model 54 4.1 Introduction . 54 4.2 The Gauss-Markov Theorem . 55 4.3 Estimation of σ2 in the GM model . 57 4.4 Implications of model selection . 60 4.4.1 Underfitting (Misspecification) . 60 4.4.2 Overfitting . 61 4.5 The Aitken model and generalized least squares . 63 5 Distributional Theory 68 5.1 Introduction . 68 i TABLE OF CONTENTS STAT 714, J. TEBBS 5.2 Multivariate normal distribution . 69 5.2.1 Probability density function . 69 5.2.2 Moment generating functions . 70 5.2.3 Properties . 72 5.2.4 Less-than-full-rank normal distributions . 73 5.2.5 Independence results . 74 5.2.6 Conditional distributions . 76 5.3 Noncentral χ2 distribution . 77 5.4 Noncentral F distribution . 79 5.5 Distributions of quadratic forms . 81 5.6 Independence of quadratic forms . 85 5.7 Cochran's Theorem .
    [Show full text]