Perfect prediction, 310
Polytomous logistic regression, 434–437
adding variables, 444–448
extending, 444–449
likelihood function for, 450–452
odds ratios from, 440
ordinal model and, 472
proportional odds model, 468–469
SAS and, 617–619
SPSS and, 643–644
Stata and, 657
Pooling, 407–409
Potential confounders, 56, 57, 60, 65,
399
Precision
confounding and, 211
consideration of, 171, 222–223
gaining, 214
interaction and, 215–223
matching and, 393
validity and, 211–212
Predicted risk, 10
Prediction, 167
Probability, 6, 18–19, 45, 112
Product terms, 28, 62, 119, 174
hierarchy principle, 190–192, 196
interaction and, 210, 399, 405
Proportional odds model, 466–472
alternate formulation of, 469
polytomous model, 468–469
Q
Quasi-likelihood
estimating equations, 522
methods, 506
R
Random effects, 579–585
Rare disease assumption, 16
Receiver operating characteristic (ROC)
curve, 349, 355–358
SAS and, 614–617
SAS and, 640–642
SAS and, 654–656
Reduced model, 54, 135
Referent group, 50, 392
Retaining variables, 185–186
Risk, 10, 43
Risk estimates, 13
Risk odds ratio (ROR), 23–24
estimated, 26
general formula for, 47–48
product formula for, 25
Risk ratio (RR), 11–13, 15–16
Robust conditions, 12
ROR.SeeRisk odds ratio
RR.SeeRisk ratio
R-to-1 matching, 393S
Sample size, 121
SAS software, 107, 553, 602–634
Saturated model, 305–307, 311
Scale factor, 522, 526–527
Score equations, 521–523
Score-like equations, 521–528
Score statistic, 140
Score test, 472–473, 480, 519–523
Screening variables, 263–270
assessing confounding, 264
assessing interaction, 264
collinearity, 264, 267
Sensitivity, 349
Simple analysis, 46–48
Single exposure variables, 45
Small samples, 121
Smallversuslarge number of
parameters debate, 108–110
Software packages.SeeComputer data
sets; specific programs
Specificity, 349
SPSS software, 107, 553, 635–648
Standard error, estimated, 140–141
Standard logistic regression, 441
Stata software, 107, 553, 649–665
Stationary m-dependent correlation
structure, 514
Statistical inference, 117–121,
130–153
Statistical tests for GEE, 519–520
Stratification, 116, 398
logistic regression, 398
matching and, 394–397
Study variables, 4
Subject-specific effects, 580–585
Subsets, eligible, 214
Symmetric matrix, 508700 Index