+- jeplus.org forums (http://jeplus.org/mybb)
+-- Forum: Building simulation tools (http://jeplus.org/mybb/forumdisplay.php?fid=1)
+--- Forum: jEPlus (http://jeplus.org/mybb/forumdisplay.php?fid=2)
+--- Thread: Relative impact of individual parameters (Solved!) (/showthread.php?tid=13)
Relative impact of individual parameters (Solved!) - oddovalspin - 10-24-2014
Hi,
I have jEPlus1.5 running successfully on my PC and have run a Latin Hypercube Sample of 150 simulations from a jEPlus job with 12 Parameters which each had 5 alternative values (potentially 244,140,625 jobs). This fulfils one of my objectives exactly. Thank you for creating this tool and your help to date.
I also have another task I need to perform. I need to assess the relative impact of each of the 12 parameters with the values of all other parameters held at a nominal value. Eventually I want to extend this to assess the relative impact of over 500 hundred parameters, but first I need to prove a method.
If I arrange my parameters in a tree with each parameter described by a low, nominal and a high value I end up with potentially 3^12 = 531,441 jobs. However, I only really need to run 3 x 12 = 36 jobs. These jobs would be:
Jobs 1 to 3: parameter1 = {low, nominal, high}, parameter2 = {nominal), parameter3 = {nominal}, parameter4 = {nominal}, etc.
Jobs 4 to 6: parameter1 = {nominal}, parameter2 = {low, nominal, high}, parameter3 = {nominal}, parameter4 = {nominal}, etc
Jobs 7 to 9 : parameter1 = {nominal}, parameter2 = {nominal}, parameter3 = {low, nominal, high}, parameter4 = {nominal}, etc
etc
Can this be configured from the Parameter Tree field on the Project Tab? How?
This would be manageable with a small number of parameters but will become unwieldy when I expand my analysis to over 500. Is there a way of configuring jEPlus from an external file such as csv or txt? I.e. A way of telling jEPlus to only run specific simulations from the tree?
Thank you in advance.
Regards, David.
RE: Relative impact of individual parameters - navid - 10-26-2014
Hi dear David, Good Question !! I Think,It can be,But,To Be Honest,I Have not done as yet and I will try it .Dear David,Why Do You Want to assess the relative impact of each of the 12 parameters with the values of all other parameters??It is a bit UnClrear for me,Do You Want to Study sensitivity analysis ??
IF YES,
I am A MATLAB Usre.I Think, it is Better to Work Both jEplus And MATLAB.I mean,You Can Run all jobs and Then Work With MATLAB and Do Every Things you Like In MATLAB.Instead Of configurring Parameter Tree field of your own choice in the Project Tab.It is Bothersome a little,but I will try To Know.for end line of your question,job list file is a way to run your specific simulations from the tree.Extcution tab/Action Part/job list in file.
Navid
RE: Relative impact of individual parameters - Yi - 10-28-2014
Hi David,
I think Navid answered it quite well - your task can be done easily using tools like Matlab. It is not worthwhile trying to create a jEPlus param tree to represent those case. Instead, you specify all parameters (their names and search tags) in a single branch tree, and then use a job list file to specify the specific combinations you want to test. More details of the job list file can be found here: http://www.jeplus.org/wiki/doku.php?id=docs:manual_1_5_advanced#job_string_and_job_list_file
Yi
RE: Relative impact of individual parameters - oddovalspin - 10-31-2014
Thanks for your help. For the benefit of anyone else reading this post, this is what I did (which worked).
I set up the parameters using the parameter tree as usual. For larger files I will do this by writing the parameter tree in excel and importing the tree using the import function.
I wrote a list of jobs in a text file using Notepad++. Each job was defined by one line which had the format:
name,0,0,value1,value2,value3,value4,value5,value6,value7,value8,value9,value10,value11,value12
;where "name" was the name I chose for the output directory, the two x "0"s nominated the first weather file and the first idf file (I am only using one of each) and value(x) is value the xth parameter in the tree.
Then I saved the file as jobslist.txt
Then I ran the simulations from the Execution tab with the radio dial checked for "job list in file" and the field browsed to my jobslist.txt file.
I can see how you could use a routine in something like Matlab or Python to generate the jobslist file. This would be convenient for large numbers of parameters. Also, by automating the process you could expect to cut down on errors such as missing commas.
I played around with running the jobs from the command line in a DOS terminal, but gave that up in favour of running the assessment from within jEPlus.
Thanks again for your help Navid and thanks again for a great tool Yi.
Yi,
You might like to add a (solved) to the end of the title for this post. I have found this helpful when searching other forums.
Regards, David.
RE: Relative impact of individual parameters - navid - 10-31-2014
Hi Dear David;My Pleasure,You Did EveryThings Completely Well.
Cheers
Navid
RE: Relative impact of individual parameters (Solved!) - Yi - 11-03-2014
Hi David,
Thanks for the suggestion. I have changed the title accordingly. I think you should be able to edit your own posts as well. Unfortunately this forum software is not as modern as stackoverflow; so have to do it manually.
Yi