HOW TO USE CLUSTERBI.FOR
************************
CLUSTERBI.FOR is a FORTRAN program which estimates the area under the
curve for each of two ROC curves (and provides the variances and covariance
needed for a Z-test of Ho:AUC1=AUC2 vs Ha:AUC1 .ne. AUC2) for a 
situation in which data is both clustered and correlated. (See WHICHPRG.TXT 
for further explanation.) For example, the same sample of patients may be
used to generate two AUC (area under the curve) estimates (i.e., the same patients may be imaged by two different modalities) and each patient may in
turn provide more than one observation (i.e., both breasts subjected to
mammography). In this case, we must account for correlation between the two
AUC estimates (because they are based on the same sample of patients) as well
as intra-cluster correlation between the observations within a single patient.

You will first need to copy the program CLUSTERBI.FOR into your home directory.

You will then need to create an input data file to run with CLUSTERBI.FOR. For
simplicity, I will refer to this file as DATA.DAT. Currently CLUSTERBI.FOR is
written for a situation in which there are 2 observations per cluster, but this
can be modified. DATA.DAT (using the example of 2 observations per cluster)
should have 2 rows of data per patient, one row for Curve 1 and the second row
for Curve 2. Each row should consist of the patient ID number (1,2,...), the
test result for obs #1, truth for obs #1 (0=normal, 1=abnormal, 9=missing), the
test result for obs #2, and truth for obs #2. The program is currently written 
to accomodate a test that yields integer values from 0-100, but this too can be
modified.

An example of what the first 4 lines of DATA.DAT might look like:
1   98  1 23  0
1   81  1 18  0
2   7   0 50  0
2   40  0 9   0

You then need to save DATA.DAT in your home directory. Change the 7th line of
the program ("open(unit=1, file=...") to specify the proper location of DATA.DAT
in your home directory, then save this edited version of CLUSTERBI.FOR.

Then you will need to issue the following statement in your command tool after
the % prompt:

source /vol/vni/ipt/bin/iptsetup.csh
f77 clusterbi.for -o clusterbi
clusterbi

You should then be prompted for the total number of patients in your datafile
and the maximum number of observations (i.e., 2) per patient. The next step, if 
all's well, will then be to open up the file CLUSTER.OUT which should be created
by CLUSTERBI and placed in your home directory. Your results will be in this
output file (i.e., area1, var1, area2, var2, covariance of the two areas, the
variance of the difference between the two areas, and the numbers of normal and
abnormal observations). Be sure you delete CLUSTER.OUT each time before you
rerun CLUSTERBI.FOR.