HOW TO USE CLUSTER.FOR
**********************
CLUSTER.FOR is a FORTRAN program which estimates the area under a single ROC
curve in a situation where the data is clustered and intra-cluster correlation
must be taken into account. For example, there may be several observations per 
patient. (See WHICHPRG.TXT for further explanation of clustered data.)

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

You will then need to create an input data file to run with CLUSTER.FOR. For
simplicity, I suggest naming this file DATA.DAT and assume that in this case 
there are 2 observations per cluster. In this case, DATA.DAT should have 2 rows 
of data per patient, one row for the first observation and one row for the 2nd.
Each row should consist of the patient ID (i.e. 1,2,...), truth (1=abnormal, 
2=normal), and the test result (the program is currently written to accomodate 
a test that takes values from 0-100).

An example of what the first 4 lines of DATA.DAT might look like:

1   2 0
1   1 80
2   1 50 
2   1 90

You then need to save DATA.DAT in your home directory. 

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

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

You should then be prompted for the total number of patients (clusters) in your
data, and then for the maximum cluster size (maximum number of observations per
patient). If the program runs properly, the file CLUSTER.OUT will be created
and should appear in your home directory. Your results will be in CLUSTER.OUT.
(Be sure to delete it each time you rerun CLUSTER.FOR.) The ROC area and its 
variance should be the 2nd and 3rd numbers listed in CLUSTER.OUT.