D. var(β 1 ) Under AN Unbalanced Design
2. var(β 1 ) For ”Reference Level” Coding
No matter how the independent variables are coded, the multilevel generalized linear model’s linearization is the same as (D.13). In this section, I will derive the variance of race which is coded as 1 for black and 0 for white veterans.The same unbalance design scenarios as discussed in Appendix D.D.2will be considered, here.
a. In RI Model
Adapting the general linearized 2-level logistic model (D.13) to the 2 level RI model, adjusted by one binary variable and coded as 1 in black and 0 for white race, yields the following results:
y∗ij = β0+ xijβ1+ µ0j+ ε∗ij (D.102)
To identify the black effect, I can rewrite the model (D.14) as the race specific models
y∗ijA = β0+ β1+ µ0j+ ε∗ij for black
yijW∗ = β0+ µ0j+ ε∗ij for white (D.103)
Therefore, the average linearized race specific response at jth site are shown as (D.104) and (D.105) for black and white, respectively.
¯ y∗.jB =
PnjB
i=1(β0+ β1+ µ0j+ ε∗ijB) njB
for black
=
PnjB
i=1(β0) +PnjB
i=1(β1) +PnjB
i=1(µ0j) +PnjB i=1(ε∗ijB) njB
= njBβ0+ njBβ1+ njBµ0j +PnjB i=1(ε∗ijB) njB
= β0+ β1+ µ0j+ ¯ε∗.jB (Eε∗.jB = 0)
= β0+ β1+ µ0j
(D.104)
¯
Similarly, we can get an average outcome as shown below.
¯
Therefore, the GLS estimate of race can be obtained by
¯
y..B∗ − ¯y..W∗ = β0+ β1− β0 = β1 (D.108)
By applying the corresponding f(H) to (D.12) and (D.11), the modified y∗ij and ε∗ij results in y∗ij = (yij− ˜πij)[˜πij(1 − ˜πij)]−1+ β0+ xijβ1
ε∗ij = εij[˜πij(1 − ˜πij)]−1 (D.109)
Then, applying Vj=σµ20 to var(Y∗ij) as (D.17) for RI model yields
var(Y∗ij) = σµ20 + [˜πij(1 − ˜πij)]−1 (D.110) Based on the linearized 2-level logistic model (D.13) and by applying the GLS estimate of β1 as (D.108), I get
I have presented yijB and yijW as (D.103). Plugging in (D.111), I will get
βˆ1 =
β0 and β1 will be constant everywhere.
βˆ1 = β0+ β1+ µ0+
As known, var(ε∗ijB) or var(ε∗ijW) are the same as σε2∗ijB and σε2∗ijW in different cells. I can write the above equation as
var( ˆβ1) = (PN
2: The related sample size from (D.21) shows as n1 = n2 = · · · = nj = n, and PN
For condition Bw¯: The related sample size from (D.21) shows as
XN
I can write (D.113) as
var( ˆβ1) = σε2∗ijB
For condition Bwj: The related sample size from (D.20) shows as
I can write (D.114) as
var( ˆβ1) = σ2∗εijB
2: The related sample size from (D.20) shows as n1 = n2 = · · · = nj = n, and PN
The related sample size from (D.20) shows as
XN
Equation (D.113) can be written as
The related sample size from (D.20) shows as
XN
b. In RC Model
Adapting the general linearized 2-level logistic model (D.13) to the 2 level RC model, yields the following rsults: adjusted by race, and coded as 1 for black and 0 fir white race.
y∗ij = β0+ xijβ1 + µ0j+ µ1j+ ε∗ij (D.124)
To identify the fixed effect of black race, I can rewrite the model (D.14) as the race specific models
yijA∗ = β0+ β1 + µ0j+ µ1j+ ε∗ij for black
y∗ijW = β0+ µ0j+ ε∗ij for white (D.125)
Therefore, the average linearized race specific responses at jth site are shown as (D.126) and (D.127) for black and white, respectively.
¯
Similarly, the average outcome yields
Therefore, the GLS estimate of race can be obtained by
¯
y..B∗ − ¯y..W∗ = β0+ β1− β0 = β1 (D.130) By applying corresponding f(H) to (D.12) and (D.11), the modified y∗ij and ε∗ij result in
y∗ij = (yij− ˜πij)[˜πij(1 − ˜πij)]−1+ β0+ xijβ1
ε∗ij = εij[˜πij(1 − ˜πij)]−1 (D.131)
Then, applying Vj=σµ20 to var(Y∗ij) as (D.17) for the RI model yields
var(Y∗ij) = σµ20 + σ2µ1 + [˜πij(1 − ˜πij)]−1 (D.132)
Based on the linearized 2-level logistic model (D.13) and by applying the GLS estimate of
I have presented yijB and yijW as (D.125). Plugging in (D.133), I will get
βˆ1 =
β0 and β1 will be constant everywhere. µ0j will be constant at jth hospital.
βˆ1 = β0+ β1+
For condition B1
2: The related sample size from (D.21) shows as n1 = n2 = · · · = nj = n, and PN
For condition Bw¯: The related sample size from (D.20) shows as
XN
I can write (D.136) as
var( ˆβ1) = σµ21
For condition Bwj: The related sample size from (D.20) shows as
XN
I can write (D.136) as
2: The related sample size from (D.20) shows as n1 = n2 = · · · = nj = n, and PN
The related sample size from (D.20) shows as
XN
Equation (D.136) can be written as
The related sample size from (D.20) shows as
XN
APPENDIX E
IMAGE OF MLWIN RESULTS USING MLWIN VERSION 2.02
Figure E1: RI Model for Pneumonia Patient Younger than 65, Using RIGLS PQL2
(a) Fixed Effect (b) Random Effect
Figure E2: Hypothesis Test for Intercept Term in the RI Model, Using RIGLS PQL2
Figure E3: Hypothesis Test for Black as Fixed Effect in the RI Model, Using RIGLS PQL2
Figure E4: RC Model for Pneumonia Patients Younger Than 65, Using RIGLS PQL2
(a) Fixed Effect (b) Random Effect
Figure E5: Hypothesis Test for Intercept Term in the RC Model, Using RIGLS PQL2
(a) Fixed Effect (b) Random Effect
Figure E6: Hypothesis Test for Black in the RC Model, Using RIGLS PQL2
Figure E7: RI Model for Pneumonia Patient Younger Than 65 in Q2Q4, Using RIGLS PQL2
(a) Fixed Effect (b) Random Effect
Figure E8: Hypothesis Test for Intercept Term in the RI Model in Q2Q4, Using RIGLS PQL2
Figure E9: Hypothesis Test for Black as Fixed Effect in the RI Model in Q2Q4, Using RIGLS PQL2
Figure E10: RC Model for Pneumonia patients younger than 65 in Q2Q4, Using RIGLS PQL2
(a) Fixed Effect (b) Random Effect
Figure E11: Hypothesis Test for Intercept Term in RC Model in Q2Q4, Using RIGLS PQL2
(a) Fixed Effect (b) Random Effect
Figure E12: Hypothesis Test for Black in RC Model in Q2Q4, Using RIGLS PQL2
Figure E13: RI Model for Pneumonia Patients Younger Than 65, Using IGLS MQL1
(a) Fixed Effect (b) Random Effect
Figure E14: Hypothesis Test for Intercept Term in the RI Model, Using IGLS MQL1
Figure E15: Hypothesis Test for Black as Fixed Effect in the RI Model, Using IGLS MQL1
Figure E16: RC Model for Pneumonia Patient Younger than 65, Using IGLS MQL1
(a) Fixed Effect (b) Random Effect
Figure E17: Hypothesis Test for Intercept Term in the RC Model, Using IGLS MQL1
(a) Fixed Effect (b) Random Effect
Figure E18: Hypothesis Test for Black in the RC Model, Using IGLS MQL1
Figure E19: RI Model for Pneumonia Patients Younger Than 65 in Q2 to Q4, Using IGLS MQL1
(a) Fixed Effect (b) Random Effect
Figure E20: Hypothesis Test for Intercept Term in RI Model in Q2 to Q4, using IGLS MQL1
Figure E21: Hypothesis Test for Black as Fixed Effect in the RI Model in Q2 to Q4, Using IGLS MQL1
Figure E22: RC Model for Pneumonia Patient Younger than 65 in Q2 to Q4, Using IGLS MQL1
(a) Fixed Effect (b) Random Effect
Figure E23: Hypothesis Test for Intercept Term in RC Model in Q2 to Q4, Using IGLS MQL1
(a) Fixed Effect (b) Random Effect
Figure E24: Hypothesis Test for Black in RC Model in Q2 to Q4, Using IGLS MQL1
APPENDIX F
PROGRAM FOR SIMULATION STUDY