These are the INSTRUCTIONS for running OBUMRM.FOR.  

1. There is important documentation at the beginning of OBUMRM.FOR
concerning what the program does, the reference to the methodology,
and the format for inputting the data.  Note that the user can
either use the fixed format specificied in the program or can
change the input format statements to accomodate their dataset.

2. There are two ways to input data into OBUMRM.FOR.  First, you
can enter the readers' confidence scores (continuous, ordinal, or
quasi-continuous).  You will get back NONPARAMETRIC estimates of
ROC AREAS and their variances and covariances.  Hypotheses about the
differences in accuracies of the diagnostic tests will be based on
these nonparametric estimates.  Sometimes, however, you may want to 
compare diagnostic tests using PARAMETRIC estimates of the ROC AREA 
or NONPARAMETRIC or PARAMETRIC estimates of an ROC index other than 
the area (e.g. partial area under the curve, or sensitivity at a 
fixed false positive rate).  In these circumstances, the user can 
input the desired estimates of accuracy for each reader in each 
diagnostic test, along with the estimated variance-covariance
matrix.  OBUMRM.FOR will calculate the test of the hypothesis that
the accuracies of the diagnostic tests are equivalent and will give
confidence intervals for the difference(s) in accuracy based on the
inputted estimates of accuracy.  

3. There are three example input files (and corresponding output files).  

i. There is one input file called "formAordin.dat" (with corresponding
output file called "formAordin_out.dat").  This gives an example of
a study comparing the accuracy of two diagnostic tests.  There are 5
readers, 69 patients without disease and 45 patients with disease.  The
readers used a five-point ordinal scale to rate their confidence, with
1 denoting "definitely no disease" and 5 denoting "definitely disease
present".  

ii. A second input file is called "formAcont.dat" (with corresponding
output file called "formAcont_out.dat").  This gives an example of
a study comparing the accuracy of two diagnostic tests.  There are 6
readers, 30 patients without disease and 30 patients with disease.  The
readers used a 0 to 100% confidence scale to rate their confidence, with
0% denoting 0% confidence in the presence of disease and 100% denoting 
100% confidence in the presence of disease.  

iii. A third input file is called "formB.dat" (with corresponding
output file called "formB_out.dat").  This gives an example of
a study comparing the accuracy of three diagnostic tests.  There are 5
readers.  The maximum likelihood estimates of the ROC area are inputted,
along with the variances and covariances.