Index 301
definition, 91
in mixture model, 109
in selection model, 107
in shared parameter model, 112
under non-future dependence, 173
missing not at random, 1, 91
definition, 94
depending on random coefficients,
112
illustration with bivariate response,
94
in mixture models, 181–206
in selection models, 167–181
mixture of bivariate normals with,
110
mixture model
applied to growth hormone trial,
234–248
applied to OASIS study, 248–261
covariate effects in, 203–206
for binary data, 194–198
hazard of dropout in, 109, 189
identification, 182–188
viaextrapolation, 187–188
viainterior family constraints, 185,
190–192
viaMAR constraints, 183–185
via non-future dependence
restrictions, 109, 186
incorporation of informative priors,
110, 222–224
applied to growth hormone trial,
234–248
applied to OASIS study, 248–261
marginalized, 215
pros and cons, 109, 112
sensitivity analysis for, 110, 190,
224–226
binary response, 196, 198
continuous response, 190, 192, 194
vs.selection models, 111, 202–203
with continuous time dropout
applied to Pediatric AIDS Trial,
261–267
with continuous-time dropout,
198–201
with discrete-time dropout, 188–198
MNAR, see missing not at random
model fit, 71, 215
posterior predictive checks, 67–68, 71
applied to CTQ II, 157–158
applied to growth hormone trial,
73–74, 146–147
under ignorability, 141–143
under nonignorability, 213–214
model selection
DIC, 63–65
applied to CTQ I, 153–154
applied to CTQ II, 157–158
applied to growth hormone trial,
73–74, 146–147
applied toHERS, 161
applied to schizophrenia trial, 150
under ignorability, 138–141
under nonignorability, 209–213
posterior predictive loss, 65–67
applied to CTQ I, 81–82
monotone missing data pattern, 89
multiple imputation, 123–124
multivariate normal model
applied to CTQ II, 155–159
applied to growth hormone trial
completers only, 72–75
ignorable, 145–149
nonignorable, 234–248
temporally aligned response, 26–27
temporally misaligned response,
27–28
multivariate probit model, 30–31, 56, 64
applied to CTQ II, 155–159
covariates in correlation matrix,
133–134
in joint models, 137–138
interpretation of parameters, 30
non-future dependence restrictions, see
mixture model
nonignorability, see missing not at
random
nonignorability index, 217
nonparametric Bayes, 68–69
applied to Pediatric AIDS Trial,
261–267