Positive Matrix Factorization Model Frequently Asked Questions

Where can I install EPA PMF?

Users can install EPA PMF anywhere they have administrative permissions. As part of the install package, the ME-2 executable, example datasets, User Guide, and supporting .dll files for calculating statistics are saved in the selected folder. The install program also creates default "EPA PMF\Data" and "EPA PMF\Output" folders in the "Libraries\Documents" folder that includes the example datasets. Most users have permissions to save files to these folders, so it is recommended that these default folders are used when operating the program.

Learn More, Register & Download:

What if I can’t install it on my hard drive (e.g. because I don’t have permissions?)

EPA PMF has reasonable performance when installed on external drives or USB drives. Runs with large datasets and error estimation runs will take longer than if run from a hard drive, but this method is typically a good choice if permissions are insufficient for the hard drive. Users must have administrative permissions on their USB drive, and not have EPA PMF already installed on the machine with the USB drive.

After installing EPA PMF I get assorted messages from Windows saying I can’t run it – what next?

Even if EPA PMF has been installed on a user’s hard drive, the user may still not have permissions to save files in the EPA PMF directory. To do this in Windows 7 or 8, users can right-click on the folder, select "properties," go to the "security" tab, and select "full control" for all users. Note that even if the user has administrative rights to a folder, these rights may not be applied to sub-folders in that folder. Depending on system set up, each folder may need to be designated with administrative rights.

On some systems, an additional step is needed by setting EPA PMF to always be "run as administrator." To do this, right click on EPA PMF 5.0.exe in the folder where it is installed and go to “properties.” Under “compatibility” or “permissions,” select "run this program as an administrator." Alternatively, the user can run as administrator by right clicking on EPA PMF 5.0.exe and selecting "run as administrator."

What format do my data need to be in?

Two input files are required by PMF:

Sample species concentration values
Sample species uncertainty values or equations for calculating uncertainty

EPA PMF accepts tab-delimited (.txt), comma-separated value (.csv), and Excel Workbook (.xls or .xlsx) files. The concentration and uncertainty files need to be the same dimensions (i.e. the same number of rows and species, with the same headers in both files).

I get a message that species names in uncertainty file do not match those in concentration file – what does this mean?

EPA PMF will still ingest the uncertainty file even if the species names do not exactly match the names in the concentration file, as long as the dimensions of the uncertainty file match the concentration file (i.e. number of rows and species).

I get a message that number of species in uncertainty file do not match number of species in concentration file – what do I do?

The concentration and uncertainty files need to have the same dimensions (i.e. the same number of rows and species), so users will need to adjust their input file to this specification outside of the program.

I can bring in data and initiate EPA PMF runs, but then get a “base run terminated before completion” message – what should I do?

Since EPA PMF is saving temporary and output files wherever it was installed, users may need to specify that they have “full control” permissions for these folders. Via Windows Explorer, right click on the folder and change permissions under “properties,” then restart EPA PMF. This should then allow EPA PMF and ME-2 to save files.

Once my data are in, what should I do first?

Look at signal/noise (S/N) ratio, summary statistics and concentration vs. uncertainty plots to evaluate input variables and identify outlier points.
If using total mass as a variable, consider making it weak.
Examine variable scatter plots to identify relationships in the data, outliers that may have a strong influence on the solution, and variables with excessive values below detection.
Make variables weak or bad based on S/N, statistics, and plot evaluation.
Run at least 20 model runs, then evaluate if solution is stable over these 20 runs (by seeing how Q varies), examine residuals and observed/predicted, and judge interpretability of factor profiles.

What if I want to save results for different analyses of the same input files?

Be sure to modify the output and configuration file names and save often as you make changes to the model run settings.

Questions/Comments

Please share EPA PMF problems and suggestions using the following form:

Submit EPA PMF Problems and Suggestions
PLEASE NOTE: EPA is tracking PMF issues and will post suggestions if many users have similar program or dataset issues. EPA does not have the ability to provide user support and many publications provide EPA PMF application examples.