Analysis in PAW
Contents:
Extracting results from a PAW (Physics Analysis Workstation) NTuple.
This WorkBook chapter is designed to introduce the Fortran-based PAW
analysis package. Although PAW has largely been superceded by ROOT, it is still very popular,
particularly among physicists who have been using the package for many
years, and also among newer workers who enjoy the quickness with which
one can navigate through ntuples.
Though based on Fortran, the PAW compiler isn't as fussy about
cases as Fortran, and will allow lower-case characters to be used.
In the Quicktour, the output of your analysis job was a ROOT file
called myHistogram.root. To obtain a HBOOK file for analysis in PAW instead,
you will need to edit your snippet.tcl file to say:
set BetaMiniTuple hbook
set histFileName myHistogram.hbook
instead of "set BetaMiniTuple root" and "set histFileName myHistogram.root."
As you can probably guess, with these settings your output will be written
to a hbook (PAW) file called myHistogram.hbook.
Now rerun your job (the commands you used in the Quicktour are shown
below, for reference):
> workdir> BetaMiniApp snippet.tcl
> mod talk KanEventInput
KanEventInput> input add /store/SP/R18/001237/200309/18.6.0b/SP_001237_013238
KanEventInput> exit
> ev beg -nev 40
> exit
(Note that you did not have to recompile or relink, because you
changed only a tcl file, not C++ code. That is one of the big advantages
of using tcl files.)
Now you should have a myHistogram.hbook in your workdir directory. A
copy of the file myHistogram.hbook from running the code above is also
available here.
You can look at HBOOK histograms with
PAW (Physics Analysis Workstation). In the session below we
will examine the output file myHistogram.hbook that
you just created. After that we will open another hbook file with
some ntuples and more histograms and extend our abilities with PAW.
You can usually invoke PAW with the simple command: paw,
which should get you to /cern/pro/bin/paw.
Start your paw session from your workdir, as that is where the output
file from the QuickTour session will be:
ana31/workdir> paw
The system should respond with something like
******************************************************
* *
* W E L C O M E to P A W *
* *
* Version 2.11/13 6 December 1999 *
* *
******************************************************
Workstation type (?=HELP) <CR>=1 :
If the environmental variable DISPLAY is correctly defined for your
session, a simple carriage-return will suffice here. If not you need
to type:
Workstation type (?=HELP) <CR>=1 : 1.aaa.bbb.ccc.ddd
where aaa.bbb.ccc.ddd is the IP address of your display.
In either event, a window called HIGSZ_01@hostname will now open
and you will get some output like:
Version 1.26/04 of HIGZ started
pawlogon.kumac executing
In private pawlogon.kumac.
PAW >
The file ~/.pawlogon.kumac would contain your own customization
of PAW, but you don't need this file to proceed. The workdir package
comes with a sample pawlogon.kumac.
PAW offers an extensive, if primitive, online help facility. When you
type "help", you get a menu of general topics.
PAW > help
From /...
1: KUIP Command Processor commands.
2: MACRO Macro Processor commands.
3: VECTOR Vector Processor commands.
4: HISTOGRAM Manipulation of histograms, Ntuples.
5: FUNCTION Operations with Functions. Creation and plotting.
6: NTUPLE Ntuple creation and related operations.
7: GRAPHICS Interface to the graphics packages HPLOT and HIGZ.
8: PICTURE Creation and manipulation of HIGZ pictures.
9: ZEBRA Interfaces to the ZEBRA RZ, FZ and DZ packages.
10: FORTRAN Interface to MINUIT, COMIS, SIGMA and
FORTRAN Input/Output.
11: NETWORK To access files on remote computers.
12: OBSOLETE Obsolete commands
Enter a number ('0'=Top, '\'=one level back, 'Q'=command mode):
Since we are interested in manipulating histograms, we choose topic 4:
Enter a number ('0'=Top, '\'=one level back, 'Q'=command mode): 4
Menu /HISTOGRAM
1: /HISTOGRAM Manipulation of histograms, Ntuples.
From /HISTOGRAM/...
2: * FILE Open an HBOOK direct access file.
3: * LIST List histograms and Ntuples in the current directory.
4: * DELETE Delete histogram/Ntuple ID in Current Directory (memory).
5: * PLOT Plot a single histogram or a 2-Dim projection.
6: * ZOOM Plot a single histogram between channels ICMIN and ICMAX.
7: * MANY_PLOTS Plot one or several histograms into the same plot.
8: * PROJECT Fill all booked projections of a 2-Dim histogram.
9: * COPY Copy a histogram onto another one.
10: * FIT Fit a user defined (and parameter dependent) function to
a histogram ID (1-Dim or 2-Dim) in the specified range.
11: 2D_PLOT Plotting of 2-Dim histograms in various formats.
12: CREATE Creation ('booking') of HBOOK objects in memory.
13: HIO Input/Output operations of histograms.
14: OPERATIONS Histogram operations and comparisons.
15: GET_VECT Fill a vector from values stored in HBOOK objects.
16: PUT_VECT Replace histogram contents with values in a vector.
17: SET Set histogram attributes.
Keyword "HISTOGRAM" found in commands:
18: * /FUNCTION/FUN1
Create a one dimensional histogram and fill the bins with
the values of a (single-valued) function.
19: * /GRAPHICS/PRIMITIVES/HIST
Draw an histogram defined by arrays X and Y.
______
0: Top level menu
Enter a number ('0'=Top, '\'=one level back, 'Q'=command mode):
Finally, we type "2" to learn about accessing the .hbook file.
The response to this comes as the output of a more for
me. I think it may be different for you depending on how your PAW
session is set up. If you get to the end of a listing and see "END"
in reverse video (where text and background switch colors), you are
probably in the more command; typing a "q" will get you
back to the PAW prompt.
Enter a number ('0'=Top, '\'=one level back, 'Q'=command mode): 2
Command "/HISTOGRAM/FILE" :
* HISTOGRAM/FILE LUN FNAME [ LRECL CHOPT ]
LUN I 'Logical unit number' R=0:128
FNAME C 'File name'
LRECL I 'Record length in words' D=1024
CHOPT C 'Options' D=' '
Possible CHOPT values are:
' ' Existing file is opened (read mode only).
N A new file is opened.
U Existing file is opened to be modified.
D Reset lock.
Open an HBOOK direct access file. If LUN is 0 the next free
logical unit will be used. If LRECL is 0 the system will
determine the correct record length of an existing file.
All this might look a little daunting, but you will only need to learn
a few simple commands to start making professional-looking histograms.
Another help feature of PAW is the usage command:
PAW > usage hist/file
* HISTOGRAM/FILE LUN FNAME [ LRECL CHOPT ]
When you open a file, you use assign that file a "logical unit
number". This allows you to work with multiple files within PAW.
So, to open our little .hbook file, we type the command:
PAW > hist/file 1 myHistogram.hbook
Note also that you can abbreviate paw commands down to unambiguous strings, so
PAW > h/fil
would also work.
Now we want to know what's in our histogram, so we study the list
command:
PAW > usage h/list
Which tells us:
* HISTOGRAM/LIST [ CHOPT ]
So enter:
PAW > h/list
Which gives us:
===> Directory :
1 (1) MC reco abs mtm difference
2 (1) Reco track momentum
4 (1) TagInspector Status
3 (1) Tracks per Event
(You may or may not have other histograms, depending on how
far you gone in the Workbook. My version of myHistogram.hbook comes
from the Quicktour stage,
so only the Tracks per Event histogram is there.
The left-most number is the id number of the histogram - this is what
you use to identify the histogram you want to plot:
PAW > usage h/plot
* HISTOGRAM/PLOT [ ID CHOPT ]
So now we can plot our histo just as we did in the quicktour:
PAW > h/pl 3
Note that if you have your HIGZ window under any other window, the
image often won't appear complete, so you need to have the HIGZ window
on top when you give the h/pl command.
As usual, to end the session:
PAW > exit
This section has discussed using PAW to view the myHistogram.hbook file
from the sample analysis job. More details on using PAW will be covered
in the following sections of this chapter.
As mentioned above, the pawlogon.kumac file contains commands that are
executed automatically when a PAW session is started. Note that the
file will only be found if the PAW session is started in the directory
where the pawlogon.kumac file is sitting.
While it is not necessary to have a pawlogon.kumac, it doesn't take
long until you realise it is actually rather nice to have one.
A few sample commands that you might like to save directly to a
pawlogon.kumac file in say, workdir, are:
* my pawlogon.kumac file
mess pawlogon.kumac executing
shell pwd
filecase KEEP
opt zfl1
These commands are:
* my pawlogon.kumac file
This is just a comment.
Any lines starting with a "*" are ignored by PAW.
mess pawlogon.kumac executing
"mess" is a message - it just prints the words
"pawlogon.kumac exiting" to the screen to tell you your
logon kumac is being run now.
shell pwd
A "shell" command is one where the user gets access to the
shell in which PAW is running and can execute normal commands. In this
case, the command is to tell us the path of the directory we are in -
in this example, when you start up paw it should tell you the path of
your workdir.
filecase KEEP
This is a good command to keep handy. PAW doesn't know about capitals
unless you execute this command. A file called MyFile.hbook
would otherwise not be able to be opened in a PAW session.
opt zfl1
This very obscure command is great - it allows you to interactively
save your pictures using the "picture" commands. More on
that later. But it's highly recommended to have this command in your
pawlogon.kumac.
kumacs I - files with multiple commands
While this section will be mostly be deferred to later in the chapter,
the basic mechanism of kumacs should be mentioned here so that if you
still have a PAW session open, you can make the pawlogon.kumac file
above and run it without having to restart PAW.
A kumac is a text file with a bunch of PAW commands. It is
common to save a frequently used collection of commands in a kumac -
for example, to do a certain nice piece of presentation or, as in the
case above, to set up to be able to use capitals, know which directory
you're in and be able to print images directly to files.
Make sure you have saved the above 6 commands into your workdir in
a file called pawlogon.kumac. You'll probably have to overwrite
a very insubstantial default pawlogon.kumac that got made the first
time you started PAW in workdir, but that's ok.
Now start a paw session, and this time after the HIGZ_01 window has
opened, a few new lines will appear in the command line window like
pawlogon.kumac executing
/afs/slac.stanford.edu/u/br/penguin/ana31/workdir
If you are in an active PAW session, you can execute any kumac,
including pawlogon.kumac by just giving the name of the file, e.g.:
exec pawlogon
where the extension is not necessary, but may also be used in the
execute command.
A larger hbook file which can used with the commands below is
available here. This is an hbook file
called pawfile.hbook which is in the same directory as this WorkBook chapter.
Now start a new PAW session and open this file:
workdir> paw
PAW > h/fil 1 pawfile.hbook
use h/lis and h/pl to have a look around the
histograms in the file - note that several are empty.
You might want to have several files open at once in PAW, for example
to compare different outputs. In the current session, we will see how
this works by opening our myHistogram.hbook file again, but this
time with a different logical unit number:
PAW > h/fil 2 myHistogram.hbook
If we now do the h/lis command, we'll see a listing of what is in the
myHistogram.hbook file - PAW has changed its focus to the most
recently-opened file. To change back to the file we opened with
logical unit number 1, we simply do:
PAW > cd //LUN1
And now an h/lis command shows us we are back with pawfile.hbook.
The numbers in brackets after the histo id number on the listing of a
hbook file are the dimension of the histos. You will see that most of
the histograms are 1D, though there are a few 2-dimensional histos
also. There are also two entries with (N) on them. These are
ntuples and will be discussed in the next section.
Ntuples are basically like spreadsheets - one row for each
event, one column for each type of information. A simple example to
help you envisage the structure of an ntuple is as follows:
Energy px py pz
event 1 5.5 .4 .4 .34
event 2 4.5 .1 -.9 .45
event 3 1 .3 .5 -.1
event 4 9 -.4 .5 .6
This ntuple has information from 4 events, and the quantities stored
are the energy and the x, y and z components of the 3-momentum.
As with normal spreadsheets, you can histogram any of these
quantities, and you can also choose to view any of these quantities
subject to constraints that that or any other quantity in the ntuple
must satisfy.
First we want to look inside one of the ntuples. Let's look at
ntuple 40, the one called "results" and see what quantities
are in the columns:
PAW > nt/print 40
Here we use "nt" since we are executing an ntuple command,
and we give the id number of the ntuple. The output includes:
* 52 * R*4 * * * DEFAULT * dauA3energycms
* 53 * R*4 * * * DEFAULT * dauB3energycms
* 54 * R*4 * * * DEFAULT * pi03energycms
* 55 * R*4 * * * DEFAULT * fourpimass
* 56 * R*4 * * * DEFAULT * omegaMass
* 57 * R*4 * * * DEFAULT * fourpimass2
* 58 * R*4 * * * DEFAULT * firstcombmass
* 59 * R*4 * * * DEFAULT * secondcombmass
* 60 * R*4 * * * DEFAULT * rhotagmass
******************************************************************
* Block * Entries * Unpacked * Packed * Packing Factor *
******************************************************************
* DEFAULT * 42 * 384 * 384 * 1.000 *
* Total * --- * 384 * 384 * 1.000 *
******************************************************************
* Blocks = 1 Variables = 60 Columns = 96 *
******************************************************************
You don't need to know much about the information here - blocks are
unimportant, there are 60 variables (that's the columns in our
illustration), and 96 columns (this is the number of events). Items
like R*4 tell you the entries are real numbers (floats).
To view information for any variable, you need to enter the ntuple
number as well as the variable name, for example:
PAW > nt/pl 40.fourpimass2
To view a 2D histogram of, say, dauA3energycms against dauB3energycms,
you can do:
PAW > nt/pl 40.dauA3energycms%dauB3energycms
In this case you'll get a fairly sparce scatter plot, but you get the
point. Note that you must be plotting quantities from the same
ntuple against each other.
It is very easy to use cuts in PAW - you just give the plot command
followed by a command stating which cut you want. Some examples are:
PAW > nt/pl 40.fourpimass omegaMass.lt.40
PAW > nt/pl 40.fourpimass fourpimass.gt.1.1
PAW > nt/pl 40.fourpimass omegaMass.lt.40.and.fourpimass.gt.1.1
PAW > nt/pl 40.fourpimass fourpimass.eq.1.2
The second line illustrates that it's ok to have decimal numbers along
with the dots around the logical operators, while the third line shows
how to use two cuts at once. Cuts can also be applied to n-dimensional
histograms.
You can also define cuts which you might use several times with the
syntax, e.g.
PAW > cuts $1 omegaMass.lt.40
PAW > cuts $2 fourpimass.gt.1.1
Then you just need to give the cut number in your plot command:
PAW > nt/pl 40.fourpimass $1.and.$2
This is a very useful way to use cuts if you are making several
histograms. Note that you can overwrite cut $N whenever you want.
In order to make use of PAW's histogram fitting abilities, ntuple
information needs to be projected onto a histogram. The procedure to
do this is as follows:
h/create/1d 100 'four [p] mass' 20 0.9 1.6
nt/proj 100 40.fourpimass
h/pl 100
The first line here creates a 1-dimensional histogram with logical
unit number (LUN) 100 (just pick a number that isn't already used),
gives it the title "four pi mass", with a greek letter to
represent pi. This histogram has 20 bins, and a range of 0.9 to 1.6.
The second line projects the column called fourpimass
of ntuple 40 onto our new histogram. Finally, the third line plots our
new histogram.
You can actually just plot the ntuple column without projecting it
onto a histogram, it's just that you don't get the full functionality
that PAW offers for histograms. It can be useful to just quickly view
the contents of a column though:
nt/plot 100 40.fourpimass
You can either define your own fit function (beyond the scope of this
WorkBook chapter) or use pre-defined fit functions provided by PAW. To
give two examples:
h/fit 100 G
h/fit 100 p6
To see fitting using a Gaussian and a 6th order polynomial, respectively.
There are two ways to plot a single-valued function with PAW: either
using a simple function, or writing a separate fortran routine
defining your function and executing that. The required syntax is:
fun/plot ufunc xlow xup [options]
where ufunc is the name of your function - either the
simple function itself or the file in which the function is defined,
xlow and xup are the lower and upper ranges
of the function argument, respectively, and the options available
are:
- C: draw a smooth curve
- S: superimpose plot on top of existing picture
- +: add contents of ID to last plotted histogram
- L: connect the points with lines
- P: draw the current marker type at each channel (bin)
- *: draw a * at each channel
A simple function can be plotted, for example:
fun/plot sin(x)/x -100 100
Whereas a more complicated function should be first defined in a text
file, for example, if we write the file FTEST.F
FUNCTION FTEST(X)
FTEST=SIN(X)*EXP(-0.1*X)
END
Making sure to use capitals to keep compatible with Fortran compilers
(though the PAW compiler isn't as fussy), and using for simplicity X
for the argument. Now you can plot your new function with the one-line
command
FUN/PLOT FTEST.F(X) 0 50
For two-dimensional plots, type
help fun2
to find out about the fun2 function for plotting 2d
functions, and the many options for plotting as surfaces, lego plots,
etc, with a range of colour options
Two-dimensional data arrays can be plotted if stored in an ascii file
of the form, say mydata.dat:
1 34
2 34
3 46
4 23
VECTOR/DELETE *
VECTOR/READ X,Y mydata.dat
GRAPH $VDIM(X,1) X Y
To plot 1-dimensional arrays, i.e. single lists of numbers, one method
is to read in the list as a vector, then put the contents of that
vector into a 1D histogram, and plot the resulting histo. To do this,
you should first determine the size of the vector - either by using
something like the command wc -l at the unix prompt on
the data file, or by using in PAW
v/print myvect
To display the contents of the vector, and thereby find out the number
of entries there are.
An example of the relevant commands in PAW to plot the list of
numbers 23,24,25,45,56 stored as a single column in the file
myvect.dat is:
v/read myvector myvec.dat
v/print myvector //look at the number of entries
//in this case it is 5
h/cr/1D 100 'vector picture' 5 0 100 //id,name,numbins,xlow,xup
h/put_vec/contents 100 myvector //put contents of myvector into
//histo with id number 100
h/pl 100 //plot as a plain histogram
h/pl 100 c s //overlay curve thru points on same plot
The options for plotting histograms are the same as those for plotting functions.
A nice histogramming function available in PAW which should be
mentioned is the many_plots function. It allows the user
to plot several histograms (up to a fairly large number) on the one
plot. The syntax is:
h/many_plots 1 2 45 67
Of course, a very useful thing would be to be able to plot several
histos on separate boxes in the same window. This is done by invoking
the zone command before plotting the images. For example:
zone 2 2
h/pl 3
h/pl 4
h/pl 9
h/pl 10
It can also be desirable to put two histos on the same plot, but with
different colours:
zone 1 1
set hcol 2
h/pl 9
set hcol 3
h/pl 10 s
set hcol 1
Here the "s" tells PAW to put the histos on the same plot.
To plot a filled-in histogram, you need to use numbers like 1002:
set hcol 1002
h/pl 9
set hcol 1
To change between log and linear axis, use:
opt liny
opt logy
opt linx
opt logx
Finally, in order to refresh the higz-window, you need type the following
command at the PAW prompt (or include it in the pawlogon.kumac):
PAW > IGSET 2buf 11
Here's where you wanted to have your
opt zfl1
command in your pawlogon.kumac file. If you don't already have that
line executed, enter it now. This line now makes it possible to write
the output you make to a file. Anything you have made before executing
this command isn't stored in the picture buffer, so can't be saved.
Having executed this command, you can save postscript or eps files
with the command:
p/print myfile.ps
p/print myfile.eps
Where the extension will be recognised by PAW and you will get the
filetype you expect.
To write to postscript files in any other way requires more knowledge
of PAW which is beyond the scope of this chapter, but can be
investigated by looking at the "file" options - namely, f/file.
Versions of PAW for Intel based platform are available from the
PAW home page.
Download all of the binaries and libraries from the CERNLIB server.
For development purposes, you also have the option to download the complete distribution.
Windows 95 users need to define the HOME environment variable in
the autoexec.bat file as follows:
set HOME=c:\tmp
Related Documents:
Download all of the binaries and libraries from the CERNLIB server.
For development purposes, you also have the option to download the
complete distribution.
Related documents:
General Related documents:
Page maintained by Adam Edwards
Last modified: January 2008
|
