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

Chapter 3 1. a. ROR¼exp(b)
b. ROR¼exp(5b)
c. ROR¼exp(2b)
d. All three estimated odds ratios should have the
same value.
e. Thebin part b is one‐fifth thebin part a; thebin
part c is one‐half thebin part a.

2. a. ROR¼exp(bþd 1 AGEþd 2 CHL)
   b. ROR¼exp(5bþ 5 d 1 AGEþ 5 d 2 CHL)
   c. ROR¼exp(2bþ 2 d 1 AGEþ 2 d 2 CHL)
   d. For a given specification of AGE and CHL, all three
   estimated odds ratios should have the same value.
   e. Thebin part b is one‐fifth thebin part a; thebin
   part c is one‐half thebin part a. The same
   relationships hold for the threed 1 s and the
   threed 2 s.


3. a. ROR¼exp(5bþ 5 d 1 AGEþ 5 d 2 SEX)
   b. ROR¼exp(bþd 1 AGEþd 2 SEX)
   c. ROR¼exp(bþd 1 AGEþd 2 SEX)
   d. For a given specification of AGE and SEX, the odds
   ratios in parts b and c should have the same value.


4. a. logit P(X)¼aþb 1 S 1 þb 2 S 2 þg 1 AGEþg 2 SEX,
   whereS 1 andS 2 are dummy variables which
   distinguish between the three SSU groupings, e.g.,
   S 1 ¼1 if low, 0 otherwise andS 2 ¼1 if medium, 0
   otherwise.
   b. Using the above dummy variables, the odds ratio is
   given by ROR¼exp(b 1 ), whereX*¼(0, 0, AGE,
   SEX) andX**¼(1, 0, AGE, SEX).
   c. logitPðXÞ¼aþb 1 S 1 þb 2 S 2 þg 1 AGEþg 2 SEX
   þd 1 ðS 1 AGEÞþd 2 ðS 1 SEXÞ
   þd 3 ðS 2 AGEÞþd 4 ðS 2 SEXÞ
   d. ROR¼exp(b 1 d 1 AGEd 2 SEX)


5. a. ROR¼exp(10b 3 )
   b. ROR¼exp(195b 1 þ 10 b 3 )


6. a. ROR¼exp(10b 3 þ 10 d 31 AGEþ 10 d 32 RACE)
   b. ROR¼expð 195 b 1 þ 10 b 3 þ 195 d 11 AGE
   þ 195 d 12 RACEþ 10 d 31 AGEþ 10 d 32 RACEÞ

Chapter 4 True‐False Questions:

1. T


2. T


3. F: unconditional


4. T


5. F: the model contains a large number of parameters


6. T

670 Test Answers