Logistic Regression: A Self-learning Text, Third Edition (Statistics in the Health Sciences)

effect on the parameter estimates or on which odds

ratios are assumed invariant. Changing the values of

the independent variables, however, may affect the

estimates of the parameters.

9. odds¼exp (a 2 þ 40 b 1 þb 2 þb 3 )


10. OR¼exp [(10)b 3 þ(10)b 4 ]¼exp (b 3 þb 4 )


11. OR¼exp (b 3 þb 4 ); the OR is the same as in Question
    10 because the odds ratio is invariant to the cut-point
    used to dichotomize the outcome categories


12. OR¼exp [(b 3 þb 4 )]

Chapter 14 True-False Questions:

1. T


2. F: there is one common correlation parameter for all
   subjects.


3. T


4. F: afunctionof the mean response is modeled as linear
   (see next question)


5. T


6. T


7. F: only the estimated standard errors of the regression
   parameter estimates are affected. The regression
   parameter estimates are unaffected


8. F: consistency is an asymptotic property (i.e., holds as
   the number of clusters approaches infinity)


9. F: the empirical variance estimator is used to estimate
   the variance of the regression parameter estimates,
   not the variance of the response variable


10. T

Chapter 15

1. Model 1: logit P (D¼ 1 jX)¼b 0 þb 1 RXþb 2 SEQUENCE
   þb 3 RXSEQ
   Model 2: logit P (D¼ 1 jX)¼b 0 þb 1 RX
   þb 2 SEQUENCE
   Model 3: logit P(D¼ 1 jX)¼b 0 þb 1 RX


2. Estimated OR (where SEQUENCE¼0)
   ¼exp(0.4184)¼1.52
   95% CI: exp[0.41841.96(0.5885)]¼(0.48, 4.82)
   Estimated OR (where SEQUENCE¼1)
   ¼exp(0.41840.2136)¼1.23
   95 %CI:exp
   ð 0 : 4184  0 : 2136 Þ

 1 : 96

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

0 : 3463 þ 0 : 6388  2 ð 0 : 3463 Þ

p

¼ð 0 : 43 ; 3 : 54 Þ

Chapter 15 687