iFit: a quick start/tutorial
- Load/import data
- Axes and Aliases
- Display objects
- Mathematical operators
- Advanced Math operators:
interpolation, catenating, fitting
- Saving your work
Commands we use in this page: iData, doc, plot, subplot,
In this document, we demonstrate shortly how to use the iFit library, which mainly manipulates iData objects. We focus on commands.
If you prefer a simpler User Interface (with menus, icons, ...),
please read the miFit help.
The object iFit is to:
Each of these tasks has dedicated, detailed, documentation, as shown
in links above.
- import transparently data files, whatever be their format [Load]
- organize the data files contents, and in particular define
their axes if needed [iData axes and
- display/plot the data [Plot]
- apply mathematical operations [Math]
- fit multi parameter models onto data sets [Fit]
- save results transparently into some common scientific formats
Currently, iFit depends on a Matlab installation, and we assume that
iFit has been installed properly before. If not so, please refer to
the Install page. Then we start Matlab,
and declare the iFit library location.
$ matlab % start Matlab (or click its icon)
We are ready to work. For the purpose of this tutorial, we shall
some 1D and 2D example data sets.
>> addpath(genpath('/path/to/iFit')) % declare the iFit library location so that we can use it
>> doc(iData) % open the iFit/iData documentation (if you feel like reading it)
Let's import a file from the iFit example
files. This is done by converting a character string, assumed
a file name, as an iData object:
>> a = iData([ ifitpath 'Data/ILL_IN6.dat' ]);
which is a neutron scattering signal by a liquid, collected by the
instrument IN6@ILL. Importation also works across the Internet with
http:// and ftp:// file specifications, as
as with compressed contents. A file selector pops-up when the file
is given as ''
string). The result of the importation is an iData
We may inquire the contents of that object with the get method
a = iData object:
[Tag] [Dimension] [Title] [Last command] [Label]
tp574029 [340 1024] 'File ILL_IN6.dat ILL Data (n...' tp574029=load(iData,...
which interestingly indicates that a Signal
has been defined (by default the largest numerical block in the
object). This Signal is an Alias
which points to the Data area of the object, and more specifically
this case to the
Data.IIIIIIIIIIIIIIIIIIIIIIIIIIIII_13 section. The naming
section has been determined automatically from the file comments.
Signal may be re-assigned any time, as well as possible axes
This common mechanism uses the Alias
The axes of the IN6 signal are related to a neutron detection time,
an angle w.r.t the direct beam. We thus define the labels for these
>> xlabel(a, 'Time channel'); % 2nd axis (columns)
which depends on the convention used to represent matrices. In
matrices are stored in column-major order, that is first axis is a
within which elements are stored along columns: Signal(row, column). As the 'X'
axis usually refers to the horizontal one for a 2D signal
(matrix), it thus relates to columns, that is the second dimension.
'Y' axis is vertical, and relates to rows, that is first dimension.
>> ylabel(a, 'Angle channel');% 1st axis (rows)
Some of the importable file formats make it easy to find the Signal,
and its corresponding axes. But other formats may contain
numerical blocks within which none appears de facto as The
Signal. This is the
same for axes. Then, the Signal and axes may be redefined in order
match the user requirements.
The concepts of Alias, Signal and Axes are of major importance,
especially as they impact the way mathematical operators apply onto
objects, and their rendering in plots.
display the object contents, that is its Signal as a function
of axes, possibly weighted by a Monitor, a single command can be
independently of the data set dimensionality.
but as, in this case, the Signal contains a single large value well
above the interesting stuff, we may simply ask to display the object
It is also possible to display more than one object overlayed onto a
single plot, or side by side in the same window.
A large set of mathematical operators can be applied on iData
and may be used to correct/analyze the Signal. Whenever the Signal
the axes are changed (by value or definition), the operator result
change accordingly. This way it is possible to define a series of
operations for data processing, independently of the data files
Let's import a second IN6 data file, and set its axes labels as for
>> b = iData([ ifitpath 'Data/ILL_IN6_2.dat' ]);
obtain the projection of this object onto the first dimension
'Angle', as a new object. Then we normalize the first object 'a' by
this projection, and plot the two initial data file contents, the
projection of the second onto the Angle axis, and the normalization
the first object by this projection.
>> xlabel(b, 'Time channel'); % 2nd axis (columns)
>> ylabel(b, 'Angle channel');% 1st axis (rows)
>> c = camproj(b)
In the latter command, the log10 applies onto the vector
the 4 objects. Also, the division is made element by element with
./ operator and the projection c
(1D signal) is extended to a 2D object
in order to perform the division.
>> subplot(log10([a b c a./c ]))
All binary operators can cope with objects that may have partly
overlapping axes, possibly with different binning, and of different
dimensionality. This is done transparently.
Also, the Error estimates associated to the Signal are
propagated, assuming distributions are Gaussians.
Advanced Math operators:
The interpolation of an object onto a new set of axes is done
transparently, whatever be the object dimensionality. An object
rebinning is obtained by interpolating the object on its own axes,
sub-set of it:
>> d = interp(a, 1/4); % rebin by dividing the length of axes by 4
a function-model onto an object is easy, in order to
determine the best parameter set that describes the Signal/Monitor. The fitting
procedure makes sure that the model matches the object
and extends it if needed.
>> d(find(d>200))=0; % first remove a large peak on the side
Any object can be cut into slices, or assembled with other objects
create a higher
dimension object. The following command puts multiplications of the
previous object side by side in order to create a 3D volume.
>> [pars,criteria,message,output] = fits(d); % fit with the default Gaussian model, extended to 2D
>> plot(log10([d output.modelValue*10]),'tight')
>> e = cat(3, [d d*2 d*10 d*5 d*2 d])
Saving your work
Any object can be exported to a large number of file formats. Some
these formats retain most of the object information, while other
formats are much more restrictive.
To save your work simply specify the object, a file name (or empty)
optionally a file format (which can also be set from the file name
>> save(a, '', 'svg')
The recommended formats are Matlab MAT, HDF5, SVG, PDF, PNG,
Matlab M and flat DAT.
- iFit/quick start -
Nov. 27, 2018 2.0.2
back to Main