IPython notebooks showing off how to use either the standalone version or CIAO version of Sherpa to fit data.
You can use the MyBinder service to run these notebooks in your browser, without having to install anything, by clicking on the "launch binder" button below. It can take a little time to start up!
WARNING this has not been updated recently so is unlikely to work.
The main differences between using the standalone version of Sherpa and that provided in CIAO are:
- CIAO comes with support for the XSPEC model library whereas you have to compile both XSPEC and Sherpa when using the standalone release (if you care about XSPEC support);
- CIAO uses Crates as its I/O libraray whereas the standalone release uses AstroPy;
- the release schedules of the two are different, so you may see slightly-different functionality in the latest released versions.
I have used the MyBinder service to create an environment that contains CIAO 4.14 which can be used to run these notebooks straight from your browser. Awesome!
We do not require Jupyter support in our conda build of CIAO but it's easy to install (we also use scipy and astropy):
> pip install jupyterlab astropy scipy
> jupyter lab
CIAO 4.16 comes with both jupyter notebook support and includes Matplotlib,
so you just need to install astropy and scipy, then you can view the
notebooks:
> pip install astropy scipy
> jupyter notebook
in this directory after starting CIAO (i.e. sourceing the ciao.*sh file
to match your shell). You can install jupyterlab with pip and use the
lab environment instead.
We provide conda packages for Sherpa (and it is also pip installable),
so you can install Sherpa into a conda environment, along with the jupyter
notebook (or lab) enviroment. For example:
> conda create -n=sherpa-notebooks -c sherpa python=3.11 sherpa jupyterlab matplotlib astropy scipy
> conda activate sherpa-notebooks
> jupyter lab
Note I am not sure if this still works (as of CIAO 4.16).
These do not have to be read in the presented order, but some build on previous ones. They are not guaranteed to be kept up to date!
-
1. A really simple "fit", which also doubles as my go-to-guide when cooking my kids porridge for breakfast.
Last updated: May 29 2024 to use Sherpa 4.16.1
-
2. A simple fit which is based on the NumPy/SciPy fitting example from the Practical Python for Astronomers course.
Last updated: May 29 2024 to use Sherpa 4.16.1
-
3. Writing your own user model, which fits the cumulative distribution function of a data set with the Gamma CDF. The same code can be run from the CIAO version of Sherpa (I added this as we are currently lacking in documentation for how to use the
add_user_modelroutine).Last updated: May 29 2024 to use Sherpa 4.16.1
-
4. An integrated user model, which fits the Gamma probability distribution function to the data set used in the 'Writing your own user model' example. This is an example of handling "integrated" data sets, and also shows off how to write a
guessroutine for your model, and some plot manipulations (in particular, re-creating a region-projection plot and adding extra annotations to it).Last updated: May 29 2024 to use Sherpa 4.16.1
-
5. How can I plot data and models when using the lower-level routines, which came up (in my mind) in the "Writing your own user model" example, where I started off using the lower-level API, but quickly switched to the higher-level (data management) API in part because I did not know how to create the plots. Well, I do know, and so can you.
Last updated: May 29 2024 to use Sherpa 4.16.1
-
6. Extending existing models (and an example of using XSPEC models), which shows how to write a user model that extends the behavior of an existing model (in this case, subtracting a model from itself with different parameter values). It also shows how to build the XSPEC module, and so use the XSPEC models from standalone Sherpa.
Last updated: January 2 2024 to use the CIAO 4.16 release
-
7. Simulating 2D data with a dash of error analysis, which uses the object API to simulate a 2D model (i.e. an image), fit it, and calculate errors on the parameters. This can be thought of as an extension of the previous notebooks that show how to replicate the functionality of the high-level UI layer using the object API (it also marks the start of me using the term "object API" for what I previously referred to as the "low-level API").
This is very slow on the myBinder environment as it needs a relatively-fast machine.
Last updated: May 29 2024 to use Sherpa 4.16.1
-
8. Simulating and fitting a 2D image (this time with a Bayesian approach), which is based on the previous notebook, this time showing how you can use the Monte Carlo Markov Chain (MCMC) analysis module in Sherpa (that is, the Bayesian Low-Count X-ray Spectral (pyBLoCXS) module). This notebook is mainly intended to show how to do this, rather than explain why (or the differences between the various frequentist and Bayesian methods for coming up with an error estimate).
This is very very slow on the myBinder environment as it needs a fast machine.
Last updated: May 29 2024 to use Sherpa 4.16.1
-
9. An introduction to the Session object, which is a very brief example of how to use the Session object to access the UI API (e.g. that from sherpa.astro.ui or sherpa.ui) but with a more-Pythonic approach.
Last updated: May 29 2024 to use Sherpa 4.16.1
-
10. Fitting a PHA dataset using the object API, which provides a quick runthrough showing how to use the object API to fit a PHA data set (that is, how in include the instrument response in a fit).
Last updated: May 29 2024 to use Sherpa 4.16.1
-
11. Grouping spectra, which looks at grouping PHA data to match plots produced by Sherpa. This isn't quite finished, but it was hopefully useful to the REXIS team.
Last updated: January 2024 to use the CIAO 4.16 release
-
Exploring a bias in WSTAT, which recreates, in our favourite fitting package, the analysis of WSTAT biases by Giacomo Vianello on the page Bias In Profile Poisson Likelihood.
-
A deep dive into PHA responses, which gets somewhat into the weeds of how Sherpa represents PHA responses (ARF and RMF, which is the trickier of the two).
-
Sherpa and the Terrible, Horrible, No Good, Very Bad Optimizer, which attempts to provide an example of how to add an optimizer to Sherpa, but because I am not much of an optimization guru, the example is defnitely not fit for human consumption.
New release: May 2024
The information in these notebooks is placed in the Public Domain and is not an official product of the Chandra X-ray Center. However, I do use these as jump-off points for adding to, or improving, the Sherpa documentation.