Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
|
|
- Brandon Riley
- 5 years ago
- Views:
Transcription
1 Biometrika Trust A note on the sensitivity to assumptions of a generalized linear mixed model Author(s): D. R. COX and M. Y. WONG Source: Biometrika, Vol. 97, No. 1 (MARCH 2010), pp Published by: Oxford University Press on behalf of Biometrika Trust Stable URL: Accessed: :37 UTC REFERENCES Linked references are available on JSTOR for this article: You may need to log in to JSTOR to access the linked references. JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org. Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at Biometrika Trust, Oxford University Press are collaborating with JSTOR to digitize, preserve and extend access to Biometrika
2 Biometrika (2010), 97, \,pp ? 2010 Biometrika Trust Printed in Great Britain doi: /biomet/asp083 Miscellanea A note on the sensitivity to assumptions of a generalized linear mixed model By D. R. COX Nuffield College, Oxford OX1 INF, U.K. david.cox@nuffield.ox.ac.uk and M. Y. WONG Department of Mathematics, Hong Kong University of Science and Technology, Clear Water Bay, Kowloon, Hong Kong mamywong@ust.hk Summary A simple case of Poisson regression is used to study the potential gain in efficiency from using a mixed model representation. Possible systematic errors arising from misspecification of the random terms in the model are examined. It is shown in particular that for a special but realistic problem, appreciable bias may arise from misspecification of a random component. Some key words: Conditional likelihood; Multilevel model; Poisson model; Recovery of information; Stratification. 1. Introduction Generalized linear mixed models are widely used, especially in the analysis of observational studies. See, for example, Breslow & Clayton (1993) and Snijders & Bosker (1999, Ch. 14). In the present paper we discuss briefly some general issues connected with such models and then examine in more detail the sensitivity to assumptions of a particular but representative special case. We also assess the nominal improvement in precision that the mixed model formulation produces. We consider for illustration a simple special situation in which observations are grouped into centres, blocks or strata. Some parameters, called structural parameters, have a common interpretation across strata and are of intrinsic interest, even if their value changes from stratum to stratum. Others are stratum-specific and are nuisance parameters for the current purpose. A random effect formulation may be used for one or both types of parameter. First we briefly discuss some of the general issues involved, but the detailed analysis in the paper is restricted to the second type of parameter. 2. Structural parameters A typical example of a structural parameter is a treatment contrast in an observational or experimental study replicated in a broadly similar and comparable form in a number of centres. If there is inexplicable variation between centres in the treatment effect, as judged by a significant treatment by centre interaction, it is common, although not entirely uncontroversial, to add a centre-specific random component to the treatment effect. For a general discussion, see Cox & Solomon (2000,? 4.5). The effect of this as compared with an assumption of constant treatment effect is, for unbalanced data, both movement of the point estimate of the treatment effect towards an unweighted average of the individual effects and increase in the notional standard error. A realistic estimation of the magnitude of the additional component of variance requires
3 210 D. R. Cox and M. Y. Wong that the number of centres is appreciable. The parameter defining the treatment effect refers to an average over an ensemble of repetitions of the stochastic system generating the interaction; it is probably rarely plausible to regard the centres as a random sample from a meaningful target population of centres, leading to an interpretation that, if available, would be more tangible. We now suppose that interest focuses on the structural parameter. 3. Formulation of a special case In order to study a specific situation, we suppose that observations are available in pairs (y7-o? J'y i ) f?r j? 1,..., m represented by independent Poisson distributed random variables with means (rjoaje'0, rjiaje6), where 0 is the structural parameter and aj is stratum-specific. Here rjo and rj\ are known constants specifying, for example, the sizes of risk sets. In this formulation the number, m, of strata may be quite large. This model is representative of, in particular, epidemiological studies of mortality in unexposed and exposed individuals stratified by, for example, age and other features. The model has been analyzed by de Stavola & Cox (2008) with the stratum parameters fixed, their objective being to examine the effects of unnecessary stratification. An essentially similar but more complicated version would allow the estimation of, say, a logistic regression equation within each stratum, some of the parameters of which are stratum-specific. 4. Some general considerations If interest is in 0, the presence of a large number of nuisance parameters is a danger signal of possible loss of sensitivity or even of inconsistency. There are several ways to proceed. The first is to replace the aj by a much smaller number of parameters. For example, if the strata are defined by age, a polynomial or other relatively simple function of age might be used. The second, and in a sense the extreme opposite, approach is to regard the ctj as totally arbitrary and to eliminate them from the likelihood by appropriate conditioning, possible in the present example. This imposes the very strong requirement that the resulting analysis should have specified properties whatever the aj may be, no matter how extreme their values or how bizarre their configuration. Another possibility, again treating the aj as arbitrary, is to calculate the profile likelihood for the parameter 0 of interest, that is, to find for each fixed 0 the maximum likelihood estimates of aj and the cor responding maximized loglikelihood function. This may lead to inconsistent estimates of 0 unless there is substantial information internally from each stratum about the corresponding o?j. In our example this would require the individual Poisson-distributed counts to be large. We do not consider this approach further. There are now three versions that involve treating the aj as random variables, typically but not necessarily values for different j being independent. The simplest, and the one studied in more detail here, is to regard the aj as independent and identically distributed random variables with a parametrically specified distribution, in particular in the present context a gamma distribution. A corresponding normal-theory issue concerns the recovery of inter-block information in unbalanced designs (Yates, 1940). Here the Uj are block effects assumed independently and identically normally distributed, an assumption given some support by randomization of treatments to blocks. Such justification is, of course, not available in observational studies. The objective here is to improve the estimation of 0, especially in situations in which in fact the stratification is largely ineffective and the lack of balance sacrifices information. Next we may regard the aj as independent and identically distributed random variables with an arbitrary distribution. Marginal maximum likelihood is now possible. For a general discussion of consistency of estimation in this formulation, see Kiefer & Wolfowitz (1956) and, for the implications for estimation in matched pair binary data, Neuhaus et al. (1994). The emphasis in this work is on showing the consistency of the resulting estimate and that indeed in some situations it is the same as that from the conditional approach, thus implying that no recovery of inter-stratum information has been achieved. Finally, it is formally possible to treat the aj as having an arbitrary joint distribution or as independent with each component having its own arbitrary distribution. It seems clear on general grounds that this cannot be distinguished from treating the aj as arbitrary unknown constants.
4 Miscellanea 211 The h-likelihood approach (Lee et al., 2006) studies predominantly inference about the stratum-specific parameters. Here we examine the parametric random effects formulation and its implications. A general point is that estimation of the dispersion of the random component of aj is somewhat akin to estimating an upper variance component from m? 1 degrees of freedom. This has very high estimation error if m is small. 5. Some likelihoods The conditional likelihood contribution for 0 from stratum j when aj is fixed and after conditioning on tj =yjo + yj\ is where dj = yjx - yj0, r,-(0) = rj0e~e + rjxee. The information about 0 is obtained from (1) as djo-tjlogrjio), (1) m l>4e?;w/0(9)' (2) 7 = 1 whereas, if it is correctly assumed that aj = a for all j, the pooled information conditionally on 5ZJ=i 0 *s ip = 4? ^?>oj (prj^j j?r'(0)- (3) Thus, in the analysis regarding the a7 as unknown constants, the analogue of the efficiency factor of an incomplete block design is (E^oHE^o.) " The behaviour of this is explored in a slightly different notation by de Stavola & Cox (2008). This quantity is not the efficiency of the conditional stratum-specific analysis that depends on the variation among the aj, but rather the loss of information in that analysis when in fact the stratification is nugatory. Now consider the likelihood when the c?j are independent and identically distributed with the gamma density ^(^a)77-1 exp(??a)/ F(r]) for a > 0. It is convenient to write v = for the reciprocal of the mean. Then the contribution of the j th stratum to the loglikelihood is where r7(#, vrj) = Vj(0) -\- vr?. log{r(i; + ri)i T{ri)} + Odj + rj log r? + rj log v - (tj + rj) logry(0, vr?), (4) It follows on differentiating twice with respect to the parameters and taking expectations over tj, dj that the parameter 77 is estimated orthogonally to 0, v and hence errors in estimating rj can be disregarded in the subsequent calculations. This corresponds in the analysis of incomplete block designs to insensitivity to the weights by which between- and within-block estimates are combined. If lm denotes the mixed model likelihood obtained by adding (4) over 7, we have that m iee = E(-d2lm/d02) = $2{4r;or,-i + j{e))i[vr?{9, vr?)}, m i0v = E(-d2lm/d0dv) = {-r1/v)y,r'j{9)lrj{e, vrj), ivv = E(-d2lm/dv2) = (n/v2)^2rj(9)/rj(e, vrj). Here r'j{6) is the derivative of rj(6) with respect to 9.
5 212 D. R. Cox and M. Y. Wong Table 1. Ratio of ic/iee.v for Coo =?ii = 1/2, 0yo = ~aj\ and different values ofro, r\ and rj r0 = 2 r0 = 10 r0 = 50 W? ri/ro Thus, the information about 0 adjusting for the estimation of v is iee.v = iee??qv/ iw The information measures given by (2) and by (3) are recovered when r? = 0 co stratum effects and when rj is very large corresponding to constant stratum eff 6. Some numerical results For numerical comparisons we may, by transformation of the rj t, for t = 0, take 0 = 0 and v = 1. Primary interest lies in how, as r? increases, the info iee.v The ratio of ic/iee.v is expressed as a function of r? and of quantities describing the pattern of variation between strata in the rjt, defined as follows. We write m mm rjt =n(l j=\ j=\ y=i Here Coo anc* c\\ specify proport sponding covariance. Numerical wo between two values?c(t. If aj? = a7o, the numbers at risk r ic = iee.v and there is no informati has aj\ =?ajo and some numerical r from the random-effects assumption 1 / V rj, is small and the two risk g In these and subsequent calculation behaviour near 0 = 00, we define 7. Sensitivity to assu A central assumption in the above in a gamma distribution. Failure of likelihood estimate 0 or to a wrong c on the former property by studyin are small biases in the gradient of th then the formal maximum likelihoo
6 Miscellanea 213 Table 2. Bias on 0 x 103 for y0 = yx = 1, ay o = O y?r a// j, aj\ divided equally between?cii <z??/ different values of r$, r\ and r? l/y/ri n/r0 r0 = 2 r0 = 10 rq = (44) 4-2(4-0) 2-5(34) 5 4-5(8-0) 3-1(4-2) 1-3(2-2) (84) 24(3-1) 0-8(1-7) (174) 11-2(10-0) 4-0(40) (15-6) 5-9(6-8) 1-6(2-2) (16-0) 3-7(4-5) 0-9(1-5) (64-1) 22-2(174) 5-1 (4-8) 5 334(354) 8-2(9-5) 1-7(2-2) (23-0) 4-5 (5-5) 0-9(1-5) The values in parentheses are bias calculated from 1000 simulations. To a first approximation we use the information matrix, /, calculated under the gamma model. Both / a (ge> gv) are proportional to the number of strata, implying that if (5) is nonnegative, the correspond estimate is inconsistent. The expected values of the gradient components are best first calculated stratum by stratum for fi values of the ocj and with 0 = 0. In fact d0 J ry(0, vrj) J 3 1 V / ry(0, vrj) J J 1 where Aj, Bj do not depend on o?y. It follows that if E((Xj) is constant, the maximum likelihood estimating equations derived from the gamma density are unbiased and, under very general conditions to this order, a consistent estimate of is obtained regardless of, for example, the distributional form of the c?j. This is not the case, however, E(c?j) depends on (ryo, rj\). To study this, suppose that E(aj) = v~l(l + Y\aji + Yoajo). (6) Then, provided that the ajt are small and we take 9 = 0, we have that ri(2r0 + vr])(yic2u + y0ci0) - r0(2rx + vr?)(yic0\ + y0cl0) *1 2 gv? Y] f /A M? * \') As a function of r/, the bias component go is greatest when r? is information to be recovered from the between-stratum variation. Table 2 shows the approximate bias in the estimate of 9 derived fr been assumed that y0 = yx = I and that either coo = 0 or c\\? coo In the symmetrical case, r\ = r0, en? coo, Yo = Y\>Q = 0, we means that 9 is consistent since i?v = 0. For a7o =? aj \, both go and set to be zero and different values of c\ \ were considered, with 1000 the simulation, 40 pairs of Poisson-distributed random variables where 9 = 0 and the aj were obtained from (6). There is very goo the simulations, shown in parentheses, and the corresponding theor the various approximations involved in the latter. The systematic err appreciable, that is when there is appreciable information in the betw the error note that the ratio of the rates in the two groups is e20, so th 0-06 in 9 is equivalent to a 13% error in the ratio of rates.
7 214 D. R. Cox and M. Y. Wong 8. Discussion We now comment briefly on the implications of these results, both for the specific situation analyzed and also more broadly We suppose that there is a parameter of interest, here 9, and a considerable number of individual parameters, here otj. There are typically two extreme estimates of 0, one assuming the otj totally arbitrary and the other assuming the aj constant. If the assumptions of the random effects model are correct, in effect a combination of the two is obtained. Table 1 shows that the gains in efficiency over the first analysis are often quite small, implying, because the estimates themselves are asymptotically efficient under appropriate assumptions, that the estimates themselves are nearly equal. In fact the random effects estimate will rarely differ appreciably from the more appropriate of the two simpler estimates given above. If, however, the assumptions of the random effects formulation are violated in that the otj are related to relevant features of the strata, then the estimate of 0 is inconsistent, in extreme cases appreciably so. This suggests the desirability of an initial check both of the need for the random effects model and of its appropriateness. First, is it reasonable to treat the structural parameter, 0, as a constant? To address this, a conditional maximum likelihood estimate of 0 can be obtained for each stratum, together with a standard error, and the resulting estimate checked for homogeneity. A more sensitive procedure is to plot the estimates against stratum-specific explanatory variables. If these consist only of the rkj then rxj and r0y may be dichotomized to form groups of strata as a basis for the comparison. If 0 is reasonably treated as a constant, estimates of the separate stratum parameters {aj} may be formed. For this there is a complementary conditional analysis for a regarding each stratum as in principle having a separate value of 0, the likelihood contribution involving a modified Bessel function. A more practicable approach is possible if 0 is assumed constant and the number of strata is large. Then we may estimate each aj from the full stratum-specific likelihood replacing 0 by its overall maximum likelihood estimate, 0, thus leading to the estimate?j = tj/rj(q). If it seems that a random-effects representation of them is both potentially useful and appropriate, then the more complex mixed model may be fitted. Appropriateness means in particular that the variation in the aj appears random. Comparison of the estimates obtained from the formal model with those obtained by more elementary procedures is desirable. Acknowledgement We thank the editor and the referees for very helpful comments. M.Y. Wong was funded by the Hong Kong Earmarked Research Grant. References Breslow, N. E. & Clayton, D. G. (1993). Approximate inference in generalized linear mixed models. J. Am. Statist Assoc. 88, Cox, D. R. & Solomon, R J. (2000). Components of Variance. Boca Raton, FL: Chapman and Hall. De Stavola, B. & Cox, D. R. (2008). On the consequences of overstratification. Biometrika 95, Kiefer, J. & Wolfowitz, J. (1956). Consistency of the maximum likelihood estimator in the presence of infinitely many nuisance parameters. Ann. Math. Statist. 27, Lee, Y., Nelder, J. A. & Pawitan, Y. (2006). Generalized Linear Models with Random Effects. Boca Raton, FL: Chapman and Hall. Neuhaus, J. M., Kalbfleisch, J. D. & Hauck, W. W. (1994). Conditions for consistent estimation in mixed-effects models for binary matched-pair data. Can. J. Statist. 22, Snijders, T. & Bosker, R. (1999). Multilevel Analysis. New York: Sage. Yates, F. (1940). The recovery of inter-block information in balanced incomplete block designs. Ann. Eugen. 10, [Received February Revised August 2009]
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
Biometrika Trust Robust Regression via Discriminant Analysis Author(s): A. C. Atkinson and D. R. Cox Source: Biometrika, Vol. 64, No. 1 (Apr., 1977), pp. 15-19 Published by: Oxford University Press on
More informationYour use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
Biometrika Trust Some Remarks on Overdispersion Author(s): D. R. Cox Source: Biometrika, Vol. 70, No. 1 (Apr., 1983), pp. 269-274 Published by: Oxford University Press on behalf of Biometrika Trust Stable
More informationYour use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
American Society for Quality A Note on the Graphical Analysis of Multidimensional Contingency Tables Author(s): D. R. Cox and Elizabeth Lauh Source: Technometrics, Vol. 9, No. 3 (Aug., 1967), pp. 481-488
More informationYour use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
A Note on the Efficiency of Least-Squares Estimates Author(s): D. R. Cox and D. V. Hinkley Source: Journal of the Royal Statistical Society. Series B (Methodological), Vol. 30, No. 2 (1968), pp. 284-289
More informationBiometrika Trust. Biometrika Trust is collaborating with JSTOR to digitize, preserve and extend access to Biometrika.
Biometrika Trust An Improved Bonferroni Procedure for Multiple Tests of Significance Author(s): R. J. Simes Source: Biometrika, Vol. 73, No. 3 (Dec., 1986), pp. 751-754 Published by: Biometrika Trust Stable
More informationFAILURE-TIME WITH DELAYED ONSET
REVSTAT Statistical Journal Volume 13 Number 3 November 2015 227 231 FAILURE-TIME WITH DELAYED ONSET Authors: Man Yu Wong Department of Mathematics Hong Kong University of Science and Technology Hong Kong
More informationYour use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
Regression Analysis when there is Prior Information about Supplementary Variables Author(s): D. R. Cox Source: Journal of the Royal Statistical Society. Series B (Methodological), Vol. 22, No. 1 (1960),
More informationYour use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
On the Estimation of the Intensity Function of a Stationary Point Process Author(s): D. R. Cox Source: Journal of the Royal Statistical Society. Series B (Methodological), Vol. 27, No. 2 (1965), pp. 332-337
More informationYour use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
Biometrika Trust Some Simple Approximate Tests for Poisson Variates Author(s): D. R. Cox Source: Biometrika, Vol. 40, No. 3/4 (Dec., 1953), pp. 354-360 Published by: Oxford University Press on behalf of
More informationAdditive and multiplicative models for the joint effect of two risk factors
Biostatistics (2005), 6, 1,pp. 1 9 doi: 10.1093/biostatistics/kxh024 Additive and multiplicative models for the joint effect of two risk factors A. BERRINGTON DE GONZÁLEZ Cancer Research UK Epidemiology
More informationThe Econometric Society is collaborating with JSTOR to digitize, preserve and extend access to Econometrica.
On the Optimal Character of the (s, S) Policy in Inventory Theory Author(s): A. Dvoretzky, J. Kiefer, J. Wolfowitz Reviewed work(s): Source: Econometrica, Vol. 21, No. 4 (Oct., 1953), pp. 586-596 Published
More informationBiometrika Trust. Biometrika Trust is collaborating with JSTOR to digitize, preserve and extend access to Biometrika.
Biometrika Trust A Stagewise Rejective Multiple Test Procedure Based on a Modified Bonferroni Test Author(s): G. Hommel Source: Biometrika, Vol. 75, No. 2 (Jun., 1988), pp. 383-386 Published by: Biometrika
More informationThe Econometric Society is collaborating with JSTOR to digitize, preserve and extend access to Econometrica.
A Set of Independent Necessary and Sufficient Conditions for Simple Majority Decision Author(s): Kenneth O. May Source: Econometrica, Vol. 20, No. 4 (Oct., 1952), pp. 680-684 Published by: The Econometric
More informationSample size calculations for logistic and Poisson regression models
Biometrika (2), 88, 4, pp. 93 99 2 Biometrika Trust Printed in Great Britain Sample size calculations for logistic and Poisson regression models BY GWOWEN SHIEH Department of Management Science, National
More informationPQL Estimation Biases in Generalized Linear Mixed Models
PQL Estimation Biases in Generalized Linear Mixed Models Woncheol Jang Johan Lim March 18, 2006 Abstract The penalized quasi-likelihood (PQL) approach is the most common estimation procedure for the generalized
More informationEach copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission.
On the Bound for a Pair of Consecutive Quartic Residues of a Prime Author(s): R. G. Bierstedt and W. H. Mills Source: Proceedings of the American Mathematical Society, Vol. 14, No. 4 (Aug., 1963), pp.
More informationDiscussion of the paper Inference for Semiparametric Models: Some Questions and an Answer by Bickel and Kwon
Discussion of the paper Inference for Semiparametric Models: Some Questions and an Answer by Bickel and Kwon Jianqing Fan Department of Statistics Chinese University of Hong Kong AND Department of Statistics
More informationIgnoring the matching variables in cohort studies - when is it valid, and why?
Ignoring the matching variables in cohort studies - when is it valid, and why? Arvid Sjölander Abstract In observational studies of the effect of an exposure on an outcome, the exposure-outcome association
More informationYour use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
Some Applications of Exponential Ordered Scores Author(s): D. R. Cox Source: Journal of the Royal Statistical Society. Series B (Methodological), Vol. 26, No. 1 (1964), pp. 103-110 Published by: Wiley
More informationEach copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission.
On Runs of Residues Author(s): D. H. Lehmer and Emma Lehmer Source: Proceedings of the American Mathematical Society, Vol. 13, No. 1 (Feb., 1962), pp. 102-106 Published by: American Mathematical Society
More informationYour use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
The Analysis of Multivariate Binary Data Author(s): D. R. Cox Source: Journal of the Royal Statistical Society. Series C (Applied Statistics), Vol. 21, No. 2 (1972), pp. 113-120 Published by: Wiley for
More informationGeneralized Linear. Mixed Models. Methods and Applications. Modern Concepts, Walter W. Stroup. Texts in Statistical Science.
Texts in Statistical Science Generalized Linear Mixed Models Modern Concepts, Methods and Applications Walter W. Stroup CRC Press Taylor & Francis Croup Boca Raton London New York CRC Press is an imprint
More informationLOGISTIC REGRESSION Joseph M. Hilbe
LOGISTIC REGRESSION Joseph M. Hilbe Arizona State University Logistic regression is the most common method used to model binary response data. When the response is binary, it typically takes the form of
More informationInternational Biometric Society is collaborating with JSTOR to digitize, preserve and extend access to Biometrics.
400: A Method for Combining Non-Independent, One-Sided Tests of Significance Author(s): Morton B. Brown Reviewed work(s): Source: Biometrics, Vol. 31, No. 4 (Dec., 1975), pp. 987-992 Published by: International
More informationYour use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
Biometrika Trust The Use of a Concomitant Variable in Selecting an Experimental Design Author(s): D. R. Cox Source: Biometrika, Vol. 44, No. 1/2 (Jun., 1957), pp. 150-158 Published by: Oxford University
More informationEach copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission.
Selecting the Better Bernoulli Treatment Using a Matched Samples Design Author(s): Ajit C. Tamhane Source: Journal of the Royal Statistical Society. Series B (Methodological), Vol. 42, No. 1 (1980), pp.
More informationUnbiased estimation of exposure odds ratios in complete records logistic regression
Unbiased estimation of exposure odds ratios in complete records logistic regression Jonathan Bartlett London School of Hygiene and Tropical Medicine www.missingdata.org.uk Centre for Statistical Methodology
More informationBiometrika Trust. Biometrika Trust is collaborating with JSTOR to digitize, preserve and extend access to Biometrika.
Biometrika Trust Discrete Sequential Boundaries for Clinical Trials Author(s): K. K. Gordon Lan and David L. DeMets Reviewed work(s): Source: Biometrika, Vol. 70, No. 3 (Dec., 1983), pp. 659-663 Published
More informationConfidence intervals for the variance component of random-effects linear models
The Stata Journal (2004) 4, Number 4, pp. 429 435 Confidence intervals for the variance component of random-effects linear models Matteo Bottai Arnold School of Public Health University of South Carolina
More informationRidge regression. Patrick Breheny. February 8. Penalized regression Ridge regression Bayesian interpretation
Patrick Breheny February 8 Patrick Breheny High-Dimensional Data Analysis (BIOS 7600) 1/27 Introduction Basic idea Standardization Large-scale testing is, of course, a big area and we could keep talking
More informationMind Association. Oxford University Press and Mind Association are collaborating with JSTOR to digitize, preserve and extend access to Mind.
Mind Association Response to Colyvan Author(s): Joseph Melia Source: Mind, New Series, Vol. 111, No. 441 (Jan., 2002), pp. 75-79 Published by: Oxford University Press on behalf of the Mind Association
More informationNew Developments in Econometrics Lecture 9: Stratified Sampling
New Developments in Econometrics Lecture 9: Stratified Sampling Jeff Wooldridge Cemmap Lectures, UCL, June 2009 1. Overview of Stratified Sampling 2. Regression Analysis 3. Clustering and Stratification
More informationA note on profile likelihood for exponential tilt mixture models
Biometrika (2009), 96, 1,pp. 229 236 C 2009 Biometrika Trust Printed in Great Britain doi: 10.1093/biomet/asn059 Advance Access publication 22 January 2009 A note on profile likelihood for exponential
More informationYour use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
Biometrika Trust Nonlinear Component of Variance Models Author(s): P. J. Solomon and D. R. Cox Source: Biometrika, Vol. 79, No. 1 (Mar., 1992), pp. 1-11 Published by: Oxford University Press on behalf
More informationarxiv: v2 [stat.me] 8 Jun 2016
Orthogonality of the Mean and Error Distribution in Generalized Linear Models 1 BY ALAN HUANG 2 and PAUL J. RATHOUZ 3 University of Technology Sydney and University of Wisconsin Madison 4th August, 2013
More informationEach copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission.
The Variance of the Product of Random Variables Author(s): Leo A. Goodman Source: Journal of the American Statistical Association, Vol. 57, No. 297 (Mar., 1962), pp. 54-60 Published by: American Statistical
More informationThe Review of Economic Studies, Ltd.
The Review of Economic Studies, Ltd. Oxford University Press http://www.jstor.org/stable/2297086. Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at.
More informationMath 423/533: The Main Theoretical Topics
Math 423/533: The Main Theoretical Topics Notation sample size n, data index i number of predictors, p (p = 2 for simple linear regression) y i : response for individual i x i = (x i1,..., x ip ) (1 p)
More informationEstimation in Generalized Linear Models with Heterogeneous Random Effects. Woncheol Jang Johan Lim. May 19, 2004
Estimation in Generalized Linear Models with Heterogeneous Random Effects Woncheol Jang Johan Lim May 19, 2004 Abstract The penalized quasi-likelihood (PQL) approach is the most common estimation procedure
More informationEach copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission.
6625 Author(s): Nicholas Strauss, Jeffrey Shallit, Don Zagier Source: The American Mathematical Monthly, Vol. 99, No. 1 (Jan., 1992), pp. 66-69 Published by: Mathematical Association of America Stable
More informationPart IV Statistics in Epidemiology
Part IV Statistics in Epidemiology There are many good statistical textbooks on the market, and we refer readers to some of these textbooks when they need statistical techniques to analyze data or to interpret
More informationNon-maximum likelihood estimation and statistical inference for linear and nonlinear mixed models
Optimum Design for Mixed Effects Non-Linear and generalized Linear Models Cambridge, August 9-12, 2011 Non-maximum likelihood estimation and statistical inference for linear and nonlinear mixed models
More informationEcological Society of America is collaborating with JSTOR to digitize, preserve and extend access to Ecology.
Measures of the Amount of Ecologic Association Between Species Author(s): Lee R. Dice Reviewed work(s): Source: Ecology, Vol. 26, No. 3 (Jul., 1945), pp. 297-302 Published by: Ecological Society of America
More informationA General Overview of Parametric Estimation and Inference Techniques.
A General Overview of Parametric Estimation and Inference Techniques. Moulinath Banerjee University of Michigan September 11, 2012 The object of statistical inference is to glean information about an underlying
More informationDetection of Influential Observation in Linear Regression. R. Dennis Cook. Technometrics, Vol. 19, No. 1. (Feb., 1977), pp
Detection of Influential Observation in Linear Regression R. Dennis Cook Technometrics, Vol. 19, No. 1. (Feb., 1977), pp. 15-18. Stable URL: http://links.jstor.org/sici?sici=0040-1706%28197702%2919%3a1%3c15%3adoioil%3e2.0.co%3b2-8
More informationNuisance parameter elimination for proportional likelihood ratio models with nonignorable missingness and random truncation
Biometrika Advance Access published October 24, 202 Biometrika (202), pp. 8 C 202 Biometrika rust Printed in Great Britain doi: 0.093/biomet/ass056 Nuisance parameter elimination for proportional likelihood
More informationFREQUENTIST BEHAVIOR OF FORMAL BAYESIAN INFERENCE
FREQUENTIST BEHAVIOR OF FORMAL BAYESIAN INFERENCE Donald A. Pierce Oregon State Univ (Emeritus), RERF Hiroshima (Retired), Oregon Health Sciences Univ (Adjunct) Ruggero Bellio Univ of Udine For Perugia
More informationPIRLS 2016 Achievement Scaling Methodology 1
CHAPTER 11 PIRLS 2016 Achievement Scaling Methodology 1 The PIRLS approach to scaling the achievement data, based on item response theory (IRT) scaling with marginal estimation, was developed originally
More informationEach copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission.
Uncountably Many Inequivalent Analytic Actions of a Compact Group on Rn Author(s): R. S. Palais and R. W. Richardson, Jr. Source: Proceedings of the American Mathematical Society, Vol. 14, No. 3 (Jun.,
More informationSurvival Analysis for Case-Cohort Studies
Survival Analysis for ase-ohort Studies Petr Klášterecký Dept. of Probability and Mathematical Statistics, Faculty of Mathematics and Physics, harles University, Prague, zech Republic e-mail: petr.klasterecky@matfyz.cz
More informationA Course in Applied Econometrics Lecture 18: Missing Data. Jeff Wooldridge IRP Lectures, UW Madison, August Linear model with IVs: y i x i u i,
A Course in Applied Econometrics Lecture 18: Missing Data Jeff Wooldridge IRP Lectures, UW Madison, August 2008 1. When Can Missing Data be Ignored? 2. Inverse Probability Weighting 3. Imputation 4. Heckman-Type
More informationBiostat 2065 Analysis of Incomplete Data
Biostat 2065 Analysis of Incomplete Data Gong Tang Dept of Biostatistics University of Pittsburgh September 13 & 15, 2005 1. Complete-case analysis (I) Complete-case analysis refers to analysis based on
More informationConstructing Ensembles of Pseudo-Experiments
Constructing Ensembles of Pseudo-Experiments Luc Demortier The Rockefeller University, New York, NY 10021, USA The frequentist interpretation of measurement results requires the specification of an ensemble
More informationGauge Plots. Gauge Plots JAPANESE BEETLE DATA MAXIMUM LIKELIHOOD FOR SPATIALLY CORRELATED DISCRETE DATA JAPANESE BEETLE DATA
JAPANESE BEETLE DATA 6 MAXIMUM LIKELIHOOD FOR SPATIALLY CORRELATED DISCRETE DATA Gauge Plots TuscaroraLisa Central Madsen Fairways, 996 January 9, 7 Grubs Adult Activity Grub Counts 6 8 Organic Matter
More informationFinite Population Sampling and Inference
Finite Population Sampling and Inference A Prediction Approach RICHARD VALLIANT ALAN H. DORFMAN RICHARD M. ROYALL A Wiley-Interscience Publication JOHN WILEY & SONS, INC. New York Chichester Weinheim Brisbane
More informationBIAS OF MAXIMUM-LIKELIHOOD ESTIMATES IN LOGISTIC AND COX REGRESSION MODELS: A COMPARATIVE SIMULATION STUDY
BIAS OF MAXIMUM-LIKELIHOOD ESTIMATES IN LOGISTIC AND COX REGRESSION MODELS: A COMPARATIVE SIMULATION STUDY Ingo Langner 1, Ralf Bender 2, Rebecca Lenz-Tönjes 1, Helmut Küchenhoff 2, Maria Blettner 2 1
More informationGROUPED SURVIVAL DATA. Florida State University and Medical College of Wisconsin
FITTING COX'S PROPORTIONAL HAZARDS MODEL USING GROUPED SURVIVAL DATA Ian W. McKeague and Mei-Jie Zhang Florida State University and Medical College of Wisconsin Cox's proportional hazard model is often
More informationCausal Hazard Ratio Estimation By Instrumental Variables or Principal Stratification. Todd MacKenzie, PhD
Causal Hazard Ratio Estimation By Instrumental Variables or Principal Stratification Todd MacKenzie, PhD Collaborators A. James O Malley Tor Tosteson Therese Stukel 2 Overview 1. Instrumental variable
More informationINFORMS is collaborating with JSTOR to digitize, preserve and extend access to Mathematics of Operations Research.
New Finite Pivoting Rules for the Simplex Method Author(s): Robert G. Bland Reviewed work(s): Source: Mathematics of Operations Research, Vol. 2, No. 2 (May, 1977), pp. 103-107 Published by: INFORMS Stable
More informationEach copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission.
Modalities in Ackermann's "Rigorous Implication" Author(s): Alan Ross Anderson and Nuel D. Belnap, Jr. Source: The Journal of Symbolic Logic, Vol. 24, No. 2 (Jun., 1959), pp. 107-111 Published by: Association
More informationThe Problem of Modeling Rare Events in ML-based Logistic Regression s Assessing Potential Remedies via MC Simulations
The Problem of Modeling Rare Events in ML-based Logistic Regression s Assessing Potential Remedies via MC Simulations Heinz Leitgöb University of Linz, Austria Problem In logistic regression, MLEs are
More informationGeneral Regression Model
Scott S. Emerson, M.D., Ph.D. Department of Biostatistics, University of Washington, Seattle, WA 98195, USA January 5, 2015 Abstract Regression analysis can be viewed as an extension of two sample statistical
More informationYour use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
A Look at Some Data on the Old Faithful Geyser Author(s): A. Azzalini and A. W. Bowman Source: Journal of the Royal Statistical Society. Series C (Applied Statistics), Vol. 39, No. 3 (1990), pp. 357-365
More informationYour use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
The Interpretation of Interaction in Contingency Tables Author(s): E. H. Simpson Source: Journal of the Royal Statistical Society. Series B (Methodological), Vol. 13, No. 2 (1951), pp. 238-241 Published
More informationSTA216: Generalized Linear Models. Lecture 1. Review and Introduction
STA216: Generalized Linear Models Lecture 1. Review and Introduction Let y 1,..., y n denote n independent observations on a response Treat y i as a realization of a random variable Y i In the general
More informationCONVERTING OBSERVED LIKELIHOOD FUNCTIONS TO TAIL PROBABILITIES. D.A.S. Fraser Mathematics Department York University North York, Ontario M3J 1P3
CONVERTING OBSERVED LIKELIHOOD FUNCTIONS TO TAIL PROBABILITIES D.A.S. Fraser Mathematics Department York University North York, Ontario M3J 1P3 N. Reid Department of Statistics University of Toronto Toronto,
More informationYour use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
A Simple Non-Desarguesian Plane Geometry Author(s): Forest Ray Moulton Source: Transactions of the American Mathematical Society, Vol. 3, No. 2 (Apr., 1902), pp. 192-195 Published by: American Mathematical
More informationCONDITIONAL LIKELIHOOD INFERENCE IN GENERALIZED LINEAR MIXED MODELS
Statistica Sinica 14(2004), 349-360 CONDITIONAL LIKELIHOOD INFERENCE IN GENERALIZED LINEAR MIXED MODELS N. Sartori and T. A. Severini University of Padova and Northwestern University Abstract: Consider
More informationOn consistency of Kendall s tau under censoring
Biometria (28), 95, 4,pp. 997 11 C 28 Biometria Trust Printed in Great Britain doi: 1.193/biomet/asn37 Advance Access publication 17 September 28 On consistency of Kendall s tau under censoring BY DAVID
More informationThe Periodogram and its Optical Analogy.
The Periodogram and Its Optical Analogy Author(s): Arthur Schuster Reviewed work(s): Source: Proceedings of the Royal Society of London. Series A, Containing Papers of a Mathematical and Physical Character,
More informationWhat if we want to estimate the mean of w from an SS sample? Let non-overlapping, exhaustive groups, W g : g 1,...G. Random
A Course in Applied Econometrics Lecture 9: tratified ampling 1. The Basic Methodology Typically, with stratified sampling, some segments of the population Jeff Wooldridge IRP Lectures, UW Madison, August
More informationThe effect of nonzero second-order interaction on combined estimators of the odds ratio
Biometrika (1978), 65, 1, pp. 191-0 Printed in Great Britain The effect of nonzero second-order interaction on combined estimators of the odds ratio BY SONJA M. MCKINLAY Department of Mathematics, Boston
More informationEmpirical Likelihood Methods for Sample Survey Data: An Overview
AUSTRIAN JOURNAL OF STATISTICS Volume 35 (2006), Number 2&3, 191 196 Empirical Likelihood Methods for Sample Survey Data: An Overview J. N. K. Rao Carleton University, Ottawa, Canada Abstract: The use
More informationMachine Learning Practice Page 2 of 2 10/28/13
Machine Learning 10-701 Practice Page 2 of 2 10/28/13 1. True or False Please give an explanation for your answer, this is worth 1 pt/question. (a) (2 points) No classifier can do better than a naive Bayes
More informationEach copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission.
On the Probability of Covering the Circle by Rom Arcs Author(s): F. W. Huffer L. A. Shepp Source: Journal of Applied Probability, Vol. 24, No. 2 (Jun., 1987), pp. 422-429 Published by: Applied Probability
More informationEach copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission.
Merging of Opinions with Increasing Information Author(s): David Blackwell and Lester Dubins Source: The Annals of Mathematical Statistics, Vol. 33, No. 3 (Sep., 1962), pp. 882-886 Published by: Institute
More informationParameter Redundancy with Covariates
Biometrika (2010), xx, x, pp. 1 9 1 2 3 4 5 6 7 C 2007 Biometrika Trust Printed in Great Britain Parameter Redundancy with Covariates By D. J. Cole and B. J. T. Morgan School of Mathematics, Statistics
More informationNon-independence in Statistical Tests for Discrete Cross-species Data
J. theor. Biol. (1997) 188, 507514 Non-independence in Statistical Tests for Discrete Cross-species Data ALAN GRAFEN* AND MARK RIDLEY * St. John s College, Oxford OX1 3JP, and the Department of Zoology,
More informationRobustness of Logit Analysis: Unobserved Heterogeneity and Misspecified Disturbances
Discussion Paper: 2006/07 Robustness of Logit Analysis: Unobserved Heterogeneity and Misspecified Disturbances J.S. Cramer www.fee.uva.nl/ke/uva-econometrics Amsterdam School of Economics Department of
More informationGeneralized Linear Models (GLZ)
Generalized Linear Models (GLZ) Generalized Linear Models (GLZ) are an extension of the linear modeling process that allows models to be fit to data that follow probability distributions other than the
More informationACE 564 Spring Lecture 8. Violations of Basic Assumptions I: Multicollinearity and Non-Sample Information. by Professor Scott H.
ACE 564 Spring 2006 Lecture 8 Violations of Basic Assumptions I: Multicollinearity and Non-Sample Information by Professor Scott H. Irwin Readings: Griffiths, Hill and Judge. "Collinear Economic Variables,
More informationMiscellanea Kernel density estimation and marginalization consistency
Biometrika (1991), 78, 2, pp. 421-5 Printed in Great Britain Miscellanea Kernel density estimation and marginalization consistency BY MIKE WEST Institute of Statistics and Decision Sciences, Duke University,
More informationThe Surprising Conditional Adventures of the Bootstrap
The Surprising Conditional Adventures of the Bootstrap G. Alastair Young Department of Mathematics Imperial College London Inaugural Lecture, 13 March 2006 Acknowledgements Early influences: Eric Renshaw,
More informationConditional Inference by Estimation of a Marginal Distribution
Conditional Inference by Estimation of a Marginal Distribution Thomas J. DiCiccio and G. Alastair Young 1 Introduction Conditional inference has been, since the seminal work of Fisher (1934), a fundamental
More informationSTATISTICAL INFERENCE FOR SURVEY DATA ANALYSIS
STATISTICAL INFERENCE FOR SURVEY DATA ANALYSIS David A Binder and Georgia R Roberts Methodology Branch, Statistics Canada, Ottawa, ON, Canada K1A 0T6 KEY WORDS: Design-based properties, Informative sampling,
More informationMohsen Pourahmadi. 1. A sampling theorem for multivariate stationary processes. J. of Multivariate Analysis, Vol. 13, No. 1 (1983),
Mohsen Pourahmadi PUBLICATIONS Books and Editorial Activities: 1. Foundations of Time Series Analysis and Prediction Theory, John Wiley, 2001. 2. Computing Science and Statistics, 31, 2000, the Proceedings
More informationTOPOLOGICAL EQUIVALENCE OF REAL BINARY FORMS
proceedings of the american mathematical society Volume 112, Number 4. August 1991 TOPOLOGICAL EQUIVALENCE OF REAL BINARY FORMS DAVID WEINBERG AND DAVE WITTE (Communicated by Frederick R. Cohen) Abstract.
More informationProportional hazards model for matched failure time data
Mathematical Statistics Stockholm University Proportional hazards model for matched failure time data Johan Zetterqvist Examensarbete 2013:1 Postal address: Mathematical Statistics Dept. of Mathematics
More informationSTAT 5500/6500 Conditional Logistic Regression for Matched Pairs
STAT 5500/6500 Conditional Logistic Regression for Matched Pairs Motivating Example: The data we will be using comes from a subset of data taken from the Los Angeles Study of the Endometrial Cancer Data
More informationPart [1.0] Measures of Classification Accuracy for the Prediction of Survival Times
Part [1.0] Measures of Classification Accuracy for the Prediction of Survival Times Patrick J. Heagerty PhD Department of Biostatistics University of Washington 1 Biomarkers Review: Cox Regression Model
More informationApproximate and Fiducial Confidence Intervals for the Difference Between Two Binomial Proportions
Approximate and Fiducial Confidence Intervals for the Difference Between Two Binomial Proportions K. Krishnamoorthy 1 and Dan Zhang University of Louisiana at Lafayette, Lafayette, LA 70504, USA SUMMARY
More informationRegression models for multivariate ordered responses via the Plackett distribution
Journal of Multivariate Analysis 99 (2008) 2472 2478 www.elsevier.com/locate/jmva Regression models for multivariate ordered responses via the Plackett distribution A. Forcina a,, V. Dardanoni b a Dipartimento
More informationPerformance Measures for Robust Design and its applications
Toshihiko Kawamura The Institute of Statistical Mathematics, Department of Data Science, Risk Analysis Research Center 10-3 Midori-cho, Tachikawa, Tokyo 190-8562, Japan kawamura@ismacjp Abstract Taguchi
More informationTable 2.14 : Distribution of 125 subjects by laboratory and +/ Category. Test Reference Laboratory Laboratory Total
2.5. Kappa Coefficient and the Paradoxes. - 31-2.5.1 Kappa s Dependency on Trait Prevalence On February 9, 2003 we received an e-mail from a researcher asking whether it would be possible to apply the
More information6.3 How the Associational Criterion Fails
6.3. HOW THE ASSOCIATIONAL CRITERION FAILS 271 is randomized. We recall that this probability can be calculated from a causal model M either directly, by simulating the intervention do( = x), or (if P
More informationAnnals of Mathematics
Annals of Mathematics The Clifford-Klein Space Forms of Indefinite Metric Author(s): Joseph A. Wolf Reviewed work(s): Source: The Annals of Mathematics, Second Series, Vol. 75, No. 1 (Jan., 1962), pp.
More informationTesting the homogeneity of variances in a two-way classification
Biomelrika (1982), 69, 2, pp. 411-6 411 Printed in Ortal Britain Testing the homogeneity of variances in a two-way classification BY G. K. SHUKLA Department of Mathematics, Indian Institute of Technology,
More informationSEQUENTIAL TESTS FOR COMPOSITE HYPOTHESES
[ 290 ] SEQUENTIAL TESTS FOR COMPOSITE HYPOTHESES BYD. R. COX Communicated by F. J. ANSCOMBE Beceived 14 August 1951 ABSTRACT. A method is given for obtaining sequential tests in the presence of nuisance
More informationProbability and Estimation. Alan Moses
Probability and Estimation Alan Moses Random variables and probability A random variable is like a variable in algebra (e.g., y=e x ), but where at least part of the variability is taken to be stochastic.
More informationInvariant HPD credible sets and MAP estimators
Bayesian Analysis (007), Number 4, pp. 681 69 Invariant HPD credible sets and MAP estimators Pierre Druilhet and Jean-Michel Marin Abstract. MAP estimators and HPD credible sets are often criticized in
More information