Each test is named diags_test_xx; for example xx = 09.  Each test refers to a python script such as diagtest09.py.
These scripts setup the inputs that will be executed in the DiagTest method execute.
To create a new one, copy one that works into a new file: for example diagtest10.py. This file needs to be modified
in order for it to execute properly.  Modify the inputs plotset, filterid, obsid, etc in such a way that  the plot set would 
run to completion without fatal errors. In the casel plot set 10, filterid = 'f_contains', obsid = 'NCEP', etc.  

In uvcdat/testing/metrics run
./diagtest10.py --datadir ~/uvcmetrics_test_data/ --baseline ~/uvcdat-testdata/baselines/metrics/ --keep True
This will create a string and execute the string; note the script should run from the command line. It would look like
diags  --outputdir /var/folders/r0/bt1s3__x5yjfntcc1_xt8qd8000696/T/tmpoOrv7K/  --model path=~/uvcmetrics_test_data/cam_output,climos=no  
    --obs path=~/uvcmetrics_test_data/obs_atmos,filter="f_contains('NCEP')",climos=yes  --package AMWG  --set 10  --var T --seasons ANN
When complete it displays the temp directory where the output of the plot set is put. Without --keep, this 
directory is deleted.  It is of the form
output directory is =  /var/folders/r0/bt1s3__x5yjfntcc1_xt8qd8000696/T/tmpFBJhqQ/
Move to this temp directory and delete those files that start with figure or end with xml; they are unnecessary for the test.
In ~/uvcdat-testdata/baselines/metrics create a directory plotset10 and copy the files from /var/folders/r0/bt1s3__x5yjfntcc1_xt8qd8000696/T/tmpFBJhqQ/
into plotset10. 

In the script plotset10 copy the name of the combined png into imagefilename and the nc file names as keys to the dictionary ncfiles.  
The values of these latter are obtained by doing ncdump -h on the file, such as ncdump -h "CLISCCP_ANN_Global_(1)_None.nc"

Once complete rerun diagtest10 without the keep flag: 
./diagtest10.py --datadir ~/uvcmetrics_test_data/ --baseline ~/uvcdat-testdata/baselines/metrics/
This just simply test to see if you get the same answers that you got when you generated the output files. 
There should be no errors once complete.

Next change CMakeLists.txt and add the new test; it has the same form as the others. Run cmake . in the build directory. 
This will pick up the new CMakeLists.txt.  Also in the build directory move to uvcdat-testdata/baselines/metrics and create 
a subdirectory plotset10. Move into this new directory and copy the files from ~/uvcdat-testdata/baselines/metrics/plotset10
into this plotset10.  Then in the build directory run ctest -V -R diags_test_15.  There should be no errors and the test passes.