Example2

Example 2:

In this example, data (data name sorbinil) from the Sorbinil Retinopathy Trial (SRT) is used. 497 type I diabetic patients who had little or no evidence of retinopathy were randomized to either Sorbinil (aldose-reductase inhibitor) or placebo. The endpoint is change in retinopathy severity level from baseline to the maximum follow-up in individual eyes. For detailed information of SRT, please check introduction of paper “Power and Sample Size Estimation for the Clustered Wilcoxon Test” (Rosner and Glynn, 2010)

The layout of sorbinil data is as follows:

id score group

1 0 1

1 10 1

2 0 2

3 10 2

3 0 2

4 0 1

4 10 1

5 10 2

5 0 2

6 10 1

6 0 1

7 0 1

8 0 1

8 10 1

9 10 2

10 0 1

11 10 2

12 0 1

12 10 1

...

score = change in retinopathy severity level;

group has values 1 and 2;

1=control group taking placebo;

2=active group taking sorbinil (an aldose-reductase).

We would like to use the data (sorbinil) from the previous study to estimate rho_hat ( ). In the new study, we proposed effect size μ=0.1 and 100 subjects for both control and active group. And we assume that all the subjects are evaluable on both eyes just as in the previous study. We want to estimate the power of new study using previous study.

(Table4 in paper by Rosner and Glynn, 2010).

The following is the syntax and output:

data sorbinil;

infile '../Balanced/sorbinil.dat';

input id 1-3 score 6-8 group 11 strat 16;

drop strat;

run;

%include "CluswilcoxPower.sas";

%CluswilcoxPower(ipnlst=100, ipmlst=100, est_mu=N, est_rho=Y,

dsn=sorbinil, varn=score, grp=group, mu=0.1,rho=,igmin=2,

igmax=2);

The output:

]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]

Previous study 11:03 Monday, July 26, 2010 8

Number of clusters of size 2 - 2..

m2=clusters of size 2 (control group)...

237

]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]

Previous study 11:03 Monday, July 26, 2010 9

Number of clusters of size 2 - 2..

n2=clusters of size 2 (active treatment group)...

241

]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]

Estimations from Previous study 11:03 Monday, July 26, 2010 10

rho_

overall

0.39113

]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]

New study 11:03 Monday, July 26, 2010 11

Number of clusters of size 1,2,3...

ipm1=clusters of size 1 (control group), ipm2=clusters of size 2 (control group)...

ipn1=clusters of size 1 (active treatment group)...

ipm2 ipn2

100 100

]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]]

New study 11:03 Monday, July 26, 2010 12

Obs w0g2 var0theta w1g2 var1theta power

1 870.238 .001149111 873.877 .001144326 0.13164