Subject Index
A
from conventional, baseline, and survivor approaches,
240
fixed-effect estimates,
214
heterogeneous patterns,
237
intraindividual growth curves,
434,
435
longitudinal trajectory,
213
model-based predictions,
237
pattern of change over time,
241
predicted values for,
228
repeated measurements,
428
six patterns/six time points,
238
time plots
two-step parametric/nonparametric mixed models, analytic results of,
468
applied to compute fixed and random effects for,
319
default integration method,
330
empirical Bayes estimate,
262
integral approximation,
263
integral of likelihood over random effects,
328
log-likelihood function, evaluation,
261
recommended approach,
263
ADL trajectory curve,
470
Age-specific transition probabilities,
409
AHEAD baseline survey,
90
survivors, percent distribution of,
239
pattern-mixture modeling to analyze,
234
TOEP covariance pattern model,
163
Akaike Information Criterion (AIC),
78,
145,
288
American Psychological Association (APA) Task Force on Statistical Inference,
27
Analysis of variance (ANOVA) method, ,
19,
207
classical model,
longitudinal data analysis,
44,
85
mean square statistics,
43
one-factor repeated measures,
41,
42
repeated-measures models,
15,
16
repeated measures,
37,
39
empirical illustration, PCL revisited, acupuncture treatment,
45
one-factor specifications,
37,
39
two-way repeated measures,
42,
43
Approximation method, empirical application of,
394
Associated missing-data indicator random vector,
460
B
Bayes formulations,
96,
448
to specify marginal mean of response probability,
319
Bayesian information criterion (BIC),
78,
145
approximation methods,
complex simulating procedures,
298
of between-subjects random effects,
363
Beck Depression Inventory-II (BDI-II) score,
53
approximation
in linear mixed models,
267
covariance estimator,
270
generalized least squares estimator,
112
linear random coefficient model,
115
model-based least squares,
217
nonlinear predictions,
116
reduced-form empirical approach,
331
retransformation method,
376
variance–covariance matrix,
112
Between-sample variations,
38
Between-subjects
Binary data, pair-wise correlations,
293
Binary longitudinal data
conditional effect, computation,
323
conventional logistic, overview of,
310
empirical illustration
marital status on probability of disability among older Americans,
327
nonlinear predictions,
335
SAS programs, analytic steps,
328
three logit models, analytic results of,
333
mixed-effects logit model, inference of,
318
random intercept logistic regression model, specification of,
313
Block design,
data structure for analysis,
86
Borrowing of strength approach,
79
C
Categorical response data, longitudinal transition models,
379
Cholesky decomposition application, on variance–covariance matrix,
176
Classical ANOVA model,
Clinical experimental studies,
26
Coding schemes, for classification factor,
147
Common variance-covariance structure,
237
Compound symmetry (CS),
135
residual covariance structure,
135
variance–covariance structure,
135
Computer programming,
426
conditional log OR on logit,
326
statistical significance,
408
on probability scale,
325
Wald chi-square statistic,
337
Confidence intervals (CI),
35
Conventional linear mixed models,
205,
208,
220
Conventional logistic, overview of,
310
Conventional odds ratio (OR),
310,
323
in longitudinal data analysis,
317
Corrected version of the AIC (AICC),
78,
145
Corresponding variance–covariance matrix,
425
mixed-effects models,
74,
417
of parameter estimates,
371
standard GEE procedure,
291
Covariance pattern model,
287
Cox model, proportional hazard rate model,
277
Cramér–Rao inequality,
480
Cross-sectional data analysis,
Cubit polynomial function,
J -shaped time trend,
82
Cumulative standard normal distribution function,
312
D
Data available, at single point of time,
Degrees of freedom (df),
39
Department of Defense (DoD),
17
Deployment Health Clinical Center (DHCC),
17
effectiveness of acupuncture treatment on PTSD,
85
longitudinal data of,
451
Dichotomous variables,
146
Disability severity score,
457,
459
conditional density function on,
457
model-based prediction of,
462
terms of joint distribution,
461
truncated linear mixed model,
459
Dispersion/scale parameter,
245
Distinctive mixture groups,
231
E
Empirical-Bayes methods
development of the retransformation method,
272
within-subject random errors in nonlinear predictions,
270,
271
Erroneous model-based predictions,
429
Error distributional function,
319,
462
Estimator of variances,
247
Exogenous latent variables,
413
Expectation-maximization (EM) algorithms,
16,
96,
106,
110
longitudinal data analysis,
108
F
Family history of diseases,
216
First order, autoregressive processes,
138
Fisher scoring algorithm,
248,
292
Fixed-effect estimates,
237
Fixed-effects multinomial logit model,
367
Flexible mixed-effects multinomial logit models,
364
Follow-up time point,
G
Gauss–Hermite abscissas,
263
Gauss–Hermite quadrature,
262
Gaussian distributions methods,
221
Generalized estimating equations (GEEs) models,
248,
281,
282
approaches
basic specifications,
282,
284
working correlation matrix,
287
empirical illustration
marital status on disability severity in older Americans,
299
marginal regression models,
295
naïve model with independence hypothesis,
282
quasi-likelihood information criteria,
288
random-effects models,
295
Generalized least squares (GLS) equation,
111
Generalized linear mixed models (GLMMs),
244
analytic convenience of,
250
between-subjects random effects,
251
complexity of data structures,
253
deriving parameter estimates,
261
empirical Bayes BLUP,
268
estimating procedures,
254
GLMs, probability distribution of,
251
hypothesis testing on fixed effects,
255
likelihood functions,
251
linearization, best linear unbiased prediction,
267
log-likelihood function,
252,
253
marginal quasi-likelihood method,
258
methods of estimating parameters,
255
adaptive Gaussian quadrature methods,
261
marginal quasi-likelihood method,
258
Markov chain Monte Carlo methods,
264
penalized quasi-likelihood (PQL) method,
256
mixed-effects logistic regression model,
273
mixed-effects multinomial logit regression models,
275
mixed-effects ordered logistic model,
274
mixed-effects Poisson regression model,
275
mixture of distributions,
255
nonlinear distribution of,
272
nonlinear predictions,
266
nonnormal distributions,
272
pseudo-response variable,
257
quasi-likelihood functions,
258
random components, retransformation,
266
variance-covariance components of,
255
retransformation method,
269
statistical inferences,
248
basic specifications,
249
hypothesis testing
procedures on fixed effects,
253
on variance components,
255
maximization procedures on fixed effects,
253
within-subject random errors in nonlinear predictions,
271
within-subject variance,
258
Generalized linear models (GLMs),
243
independence hypothesis,
248
log-likelihood function,
247
maximum likelihood estimator (MLE),
247,
346
probability distribution,
245
regression coefficient of covariate,
246
statistical expression,
246
statistical inferences,
245
systematic component,
246
user-friendly procedures,
312
variance–covariance matrix for estimates,
248
Genetic predisposition,
216
conditional independence hypothesis,
423
individual-level likelihood function,
424
LGMM, statistical perspective,
422,
439
longitudinal sequence of response measurement,
422
multinomial function as,
423
GROUP BY TIME option,
223
H
Handling missing data,
443
Heckman’s classical two-step estimator,
472
Hessian of log-likelihood,
247
Heterogeneous linear mixed model,
218
Heterogeneous mixture patterns, classification of,
230
Heterogeneous transition pattern,
386
Hotelling’s trace test statistic,
50,
52
Household and Retirement Survey (HRS),
18
Hybrid variance–covariance structure,
142
Hypothesis testing, on nonnegative variance,
145
I
ID number, ,
Independence from irrelevant alternatives (IIA) hypothesis,
346
Individual-level likelihood function,
424
Influence diagnostics,
181
Cook’s distance statistic,
181
empirical illustrations,
190
linear mixed model concerning marital status/disability severity among older Americans,
198
PCL score, acupuncture treatment
linear mixed model, checks,
190
likelihood displacement statistic approximation,
187
LMAX statistic, for influential observations identification,
189
Institute for Social Research (ISR),
18
Integral approximation methods,
255
Intraclass correlation (ICC), ,
70
Intraindividual correlation (IIC),
10,
67,
70,
316
Intraindividual growth patterns,
20
Irregular time trends,
85
L
Laplace approximation,
255
Last observation carried forward (LOCF) approach,
444
classical method handling missing data,
446
Latent endogenous random variables,
413
Latent growth curve model (LGCM),
230,
411
Latent growth mixture model (LGMM),
230,
411,
419
latent growth modeling,
411
maximum likelihood approach,
420
Latent growth model (LGM),
411,
416
assumption of multivariate normality,
418
empirical illustration, marital status effect on ADL count,
425
factor-scoring coefficients,
417
intraindividual growth curves,
434,
435
linear slope component,
417
structural equation modeling, overview of,
412
Latent variable model,
413
structural equation model construction,
413
Leverage measurement,
183
Likelihood-based method,
79
canonical parameters,
283
for checking the polynomial form of time,
79
ADL count, for three linear mixed models,
201
Likelihood ratio statistic,
116
asymptotic null distribution of,
116
Gaussian quadrature use of,
263
goodness-of-fit information,
333
log-likelihood function,
77
Linear discriminant analysis (LDA)
random-effects regression models,
230
Linearization-based approaches,
278
Linearization methods,
259
empirical illustrations
baseline score, adjustment,
210
BLUPs
vs. least squares means,
222
marital status and disability severity in older Americans,
89
PCL score, acupuncture treatment,
85
fixed effects, inference/estimation of,
73
maximum likelihood methods,
73
statistical/hypothesis testing,
75
variance–covariance components,
71
general specification of,
67
intraindividual correlation,
69
longitudinal data analysis
baseline response, adjustment,
206
baseline score, adjustment,
206,
208
one-factor with random intercept,
62
pattern-mixture modeling,
229,
232
empirical illustration of,
234
heterogeneous groups, classification,
229
random effects
assumed distribution, misspecification,
216
in different distributions,
220
heterogeneity linear mixed model,
217
nonnormal random effect distribution,
218
random intercept and random slope,
64
polynomial time functions,
80
reduce collinearity methods,
82
variance–covariance matrix,
69
Linear random coefficient model,
218
Linear random intercept model,
220
Linear regression estimator,
461
Linear regression model,
14
Log-gamma distributed random coefficient model,
220
Log-gamma distributed slopes,
219
Log-gamma linear mixed model,
219
Logistic cumulative distribution function,
274
Logistic regression model,
380,
445
Log-likelihood functions,
73,
74,
107,
187,
219,
247,
311,
317,
319,
346,
351,
418
expression, complete-data,
421
Log-linear probability distribution,
300
Longitudinal clinical controlled trial,
442
Longitudinal courses,
ANOVA, repeated measures of,
37
balanced/unbalanced, ,
book/data organization, for illustrations,
16
asset and health dynamics among the oldest old (AHEAD),
18
confidence interval, effect size estimators,
23,
24
defined,
empirical illustration,
29
history of,
intraindividual correlation,
14
MANOVA, repeated measures of,
47
missing data, patterns and mechanisms,
monotone missing pattern,
nonmonotone missing data,
PTSD symptom, acupuncture treatment effectiveness,
29
randomized controlled clinical trial,
17
time plots, of trends,
20
time scale/number of time points,
12
Longitudinal data designs,
applications, statistical techniques,
risk factors over time points,
unbalanced,
Longitudinal data structures,
balanced/unbalanced,
multivariate data,
univariate data,
Longitudinal modeling, basic expressions of,
13
Longitudinal processes, sources of correlation,
10
Longitudinal regression models, on nonignorable missing data,
456
Longitudinal trajectories
of mortality derived,
377
for probability of disability prediction,
339
Longitudinal transition models
for categorical response data,
379
empirical illustration
measures/models/SAS programs,
395
predicted transition probabilities in functional status/marital status,
395
transition probabilities
effects of marital status,
407
mixed-effects multinomial logit transition model,
386
statistical inference,
390
variance–covariance matrix approximation for transition probabilities,
392
with only fixed effects,
384
two-time multinomial transition modeling, overview of,
380
M
discrete probability change,
325,
356
variance-covariance matrix for,
241
Marginal predictions,
240
Marginal quasi-likelihood (MQL) technique,
255,
263,
282,
352
Marginal regression model,
286
Marital status
conditional effects of,
369,
408
on probability of disability among older Americans,
327
on transition probabilities,
408
approximation method,
392
imputation on variable PCL_SUM,
452
Markov chain process,
264
fixed-effects techniques,
379
response at time point,
389
Markov random variable,
384
Maximized log-likelihoods,
77
Maximum likelihood (ML),
179,
246
estimates
factor-scoring coefficients,
430
log-likelihood functions,
107
Fisher information matrix,
248
for three linear mixed models on ADL count,
201
log-likelihood function,
260
solution for regression coefficients,
306
Mean PCL scores, time plot of,
33
Mean square (MS)
estimated effect size,
26–27
medical treatment on PTSD,
26
Methods handling missing data
mixed-effects regression models,
441
empirical illustration
analytic results/ADL count predictions,
468
pattern of change over time in ADL count,
470
measures/models/SAS programming,
463
nonignorable missing data,
463
impact of nonignorable missing data,
456
nonparametric regression model, on nonignorable missing data,
461
pattern mixture model on MNAR,
460
selection model on MNAR,
458
empirical illustration, analytic results with and without multiple imputations,
451
last observation carried forward (LOCF),
446
multiple imputations (MI),
447
and shrinkage, comparison,
450
Metropolis–Hastings algorithm,
265
Minimum mean square error, of prediction,
220
mathematical definitions of,
441
missing-data mechanism,
444
Missing completely at random (MCAR),
10,
442
longitudinal data analysis,
443
mathematical definitions of,
441
classification, natural extension of,
230
patterns,
classification standards for,
229
monotone missing pattern,
nonmonotone missing data,
patterns and mechanisms,
well-assumed prior distribution of,
448
Missing not at random (MNAR),
10,
442
mathematical definitions of,
441
missing-data mechanism,
442,
443
nonignorable missing data,
444
analytic results/nonlinear predictions,
364
analytic steps with SAS programs,
359
data, measures, and models,
358
graphical analysis, on nonlinear predictions,
374
marital status
conditional effects of,
369
longitudinal trajectories of disability/mortality,
357
approximation method, empirical application of,
394
covariates’ conditional effects on probability scale,
355
fixed/random effects, estimation of,
351
longitudinal data analysis,
343
and nonlinear predictions,
347
variance–covariance matrix approximation on probabilities,
353
ordered logistic model,
274
Poisson regression model,
275
Model-based longitudinal trajectory,
472
Modeling nonignorable missing data,
471
Modeling normal longitudinal data,
150
marginal means of
approximate variance–covariance matrix,
401
covariates’ conditional effects on probability scale,
355
empirical illustration
analytic results/nonlinear predictions,
364
analytic steps with SAS programs,
359
data, measures, and models,
358
graphical analysis, on nonlinear predictions,
374
marital status, conditional effects of,
369
marital status/longitudinal trajectories of disability/mortality,
357
fixed/random effects, estimation of,
351
longitudinal data analysis,
343
and nonlinear predictions,
347
variance–covariance matrix approximation on probabilities,
353
regression model
likelihood function for,
345
Multiple imputations (MI),
444,
447
handling missing data,
450
and shrinkage, comparison,
450
Multivariate analysis of variance (MANOVA),
19,
47
constant error variance–covariance matrix,
51
distinctive disadvantages,
49
empirical illustration
psychiatric disorders, acupuncture treatment,
53
repeated measures,
47,
51
total sums of squares,
47
Wilks’ lambda distribution,
49
Multivariate data
of repeated measurements,
structure,
distinctive disadvantages,
vs. univariate longitudinal data matrix,
N
Naïve variance estimator, in longitudinal data analysis,
284
National Death Index (NDI),
18
Newton-Raphson (NR)
Nonignorable missing data,
472
clinical experimental studies,
455
in longitudinal data analysis,
456,
471
multivariate regression models,
472
nonparametric regression model,
461
selection model to handle,
472
statistical methods for handling,
115,
444
Nonlinear longitudinal data, generalized linear models,
243
Nonlinear predictions,
347
mixed-effects multinomial logit model,
347
model-based and empirical BLUPs,
229,
267
probability of disability,
335
of random components,
266
retransformation method,
339
transition probabilities,
391
within-subject random errors,
270
Nonmonotone missing-data pattern,
450
Nonparametric mixed-effects model,
462
Nonrandom factor-scoring coefficients,
431
Null hypothesis (Type-I error),
37
O
Odds ratios (OR)
mixed-effects logit model,
312,
337
probability of disability and conditional,
338
ODS OUTPUT statement,
157
convenient curvilinear expression,
475
least-square estimators,
476
P
Panel data,
Parameter estimates, asymptotic standard errors of,
419
Parametric hazard regression model,
277
conditional distribution,
232
individual pattern indicator,
233
pattern-specific parameter estimates,
233
Patterson’s expression,
105
Pearson-type residuals,
175
approximation techniques,
255
pseudo-likelihood estimates of model parameters,
256
Physiological senescence parameters,
216
Pillai’s trace statistic,
50,
52
Poisson distribution,
276,
423
Polynomial time function, longitudinal data analysis,
84
Posttraumatic stress disorder (PTSD) symptom,
29
acupuncture treatment,
153
randomized controlled clinical trial,
17
Practical significance,
24
Predicted probabilities, of disability/death,
368
Predicted response probability, variance approximation,
320
Predicted transition probabilities,
405,
407
in functional status and marital status,
395
variance approximates,
404
variance–covariance matrix for,
380,
382,
386
Prentice’s expansion,
290
Pre–post paired
t-test,
21
PRESS statistic, defined,
183
Probability density function,
261
Probability of correct model,
424
Probit regression models, overview of,
310
Probit survival model,
458
PROC CALIS procedure,
428,
430
PROC GENMOD procedure models,
301
multinomial distribution,
360
PROC NLMIXED procedure in SAS system,
359
random effect parameters,
263
PROC LOGISTIC statement,
306,
465
PROC MIXED procedures,
87,
115,
118,
119,
125,
140,
156,
163,
198,
225,
235,
453,
467,
468
estimates regression coefficients,
361
in SAS is applied to yield parameter estimation,
226
Proportional odds assumption,
274
Pseudo-likelihood estimates,
260
acupuncture treatment,
155,
212
fixed effects of three linear mixed models,
212
pattern of change over time,
159
prediction
subject-specific time plot,
31,
123
Q
Q -point Gaussian quadrature rule,
263
Quadratic polynomial function,
80,
82
high-order polynomial functions,
82
R
Random coefficient model,
179,
180
logistic regression model, specification of,
316
model specification/SAS program,
483–485
AHEAD longitudinal data,
485
multinomial logit model on health,
483
Random-effects multinomial logit transition model, maximum likelihood estimates,
392
covariance structure for within-subject,
133
variance–covariance matrix,
133
Random intercept linear model,
176
Random intercept logit model,
313,
316
empirical BLUP approach and retransformation method based on,
328
regression model, specification of,
313
within-subject error,
315
within-subject variability,
314
Random intercept model,
314
Random intercept regression model,
350
multinomial logit models,
366,
396
Randomized block design,
39
Randomized controlled clinical trials,
208,
446
Reduced-form fixed-effects multinomial logit model,
364
Regression coefficients,
355,
454
log-likelihood function,
247
on longitudinal data,
281
multinomial logit regression modeling,
391
standardized diagnostic method,
189
Repeated measures ANOVA,
39
Residual covariance structure, patterns,
133
between-subjects variance component,
133
classification factor
coding schemes of time,
149
empirical illustrations
linear regression model,
153
marital status/disability severity among older Americans,
161
PCL score, acupuncture treatment,
161
two linear regression models, estimation,
153
autoregressive structures (AR),
137
compound symmetry (CS),
135
toeplitz structures (TOEP),
138
unstructured pattern (UN),
136
nonzero off-diagonal elements,
134
with unequal time intervals,
139
hybrid residual covariance model,
142
spatial exponential model,
141
spatial Gaussian pattern model SP(GAU),
141
variance–covariance pattern models,
134
Residual diagnostics,
174
semivariogram
in linear random coefficient model,
178
in random intercept linear models,
176
Residual log-likelihood function, maximization,
103
Residual variance–covariance matrix,
70,
232
Restricted maximum likelihood (REML) approach,
75,
95
Bayesian inference, overview of,
96
computational procedures,
106
estimators,
96,
99,
101,
102,
105,
106,
117,
120,
126,
151,
175,
256
Expectation–Maximization (EM) algorithm,
108
hypothesis testing, on variance component
G,
116
linear mixed models, estimator,
102
approximation of random effects,
111
best linear unbiased prediction (BLUP),
111
empirical illustrations,
117
marital status and disability among older americans,
124
PCL score, acupuncture treatment on,
117
shrinkage/reliability,
113
MLE bias, in variance estimate,
99
ML estimators, comparison,
105
Newton–Raphson (NR) algorithm,
107
REML, in general linear models,
101
Retransformation method,
357
based on random intercept multinomial logit model,
362
Bayesian inference, development,
272
longitudinal trajectories of health probabilities,
359
mean multinomial logit functions,
396
mixed-effects multinomial logit model,
344
transition probabilities,
405
Retransformation, mixed-effects logistic regression model,
273
Root of mean square error (RMSE),
191
Roy’s greatest root criterion,
50,
52
S
SAS PROC GLIMMIX procedure,
328
SAS PROC SGPLOT steps,
30
SAS PROC TRAJ algorithm,
425
SAS programming,
32,
55,
56,
58,
86,
87,
91,
118,
120,
122,
153,
154,
158,
162,
168,
191,
210,
211,
223–226,
235,
236,
300,
302–304,
306,
307,
330,
360–362,
370,
397,
399,
400,
432,
483
covariance parameter,
202
reduced-form random intercept multinomial logit model,
363
SAS system, PROC NLMIXED procedure,
359
Satterthwaite approximation,
77,
449
Semi-Markov transition process,
389
in linear random coefficient model,
178
in random intercept linear models,
176,
178
Spatial Gaussian covariance pattern model,
141
Spatial Gaussian pattern model (SPGAU),
141
Square root of approximated variance,
322
Standard deviation (SD),
35
Statistical models handling missing data,
442
Stochastic missing-data indicator matrix,
231
Stochastic variations, decomposition of,
177
Structural equation modeling (SEM),
411,
430
basic null hypothesis,
415
estimating procedure,
414
maximum likelihood function,
415
types of random variables,
412
Structural equation modeling, overview of,
412,
426
Subject-specific random effects,
12
Subject-specific regression coefficients,
314
T
Taylor series approximation,
321,
354
Taylor series expansion,
298,
353
Thompson’s expression,
105
Time-independent random parameter,
314
Time trends
in predicted probability of disability,
375
Time-varying covariates,
TOEP covariance structure,
144
TOEP residual variance–covariance pattern model,
164
TOEP variance–covariance structure,
164
Transition models, two-time transition models,
380
Transition probabilities,
406
TREAT-by-TIME interaction,
57
U
Univariate longitudinal data,
Unknown distribution,
462
V
Variance-component models,
Variance-covariance matrix,
72,
477
generalized linear mixed models (GLMMs),
243
within-subjects random errors,
387
W
Walter Reed National Military Medical Center (WRNMMC),
17
Weibull distributional function,
277
Wide table format,
Wilks’ lambda distribution,
49,
50,
52
Within-sample variations,
38
component, random errors,
38
variations of sample data,
38
Within-study effect sizes,
28
local approximations for,
390
multinomial logit regression modeling,
391
probabilities at series of time points,
349
variance-covariance matrix of,
387
Within-subject variability,
314,
348
Z
Zero-inflated Poisson (ZIP) distribution,
423