README

SuzUtils - Suzanne Aigrain

This repository contains miscellaneous bits of code I have written or
pilfered, which I use often, and which I thought might be useful to
others, but weren't substantial enough to warrant their own
repository. I have tried to document each routine / module reasonably
well. Comments / feedback welcome! Also please let me know if you make
significant improvements to any of these that you think I should
incorporate.

Currently included (last updated December 2013)

- periodograms: Various tools for time-series analysis in the Fourier domain.
  - sinefit: least squares fitting of sine curves to a time-series
    with arbitrary time-sampling (this is a generalisation of the
    Lomb-Scargle periodogram, but it doesn't include the false alarm
    probability and significance measures, which are almost always
    misleading anyway)
  - DftPowerspectrum: compute and plot the power spectrum of a regularly
    sampled time-series using the discrete FFT (shows how to get the
    normalisation right, and which frequencies correspond to which
    element in the array)
  - AcfPeriodogram: compute and plot the ACF of a regularly sampled
    time-series and its amplitude spectrum (also known as the
    periodogram of the time-series), after optional smoothing.

- filter: routines to filter time-series (or any 1-D dataset)
  - filt1d: combination of linear (boxcar or gauss) and median filter,
    with edge reflection or circular edge criterion
  - boxcare: boxcar (moving average) filter with edge reflection or
    circular edge criterion
  - gausse: Gaussian (tapered window) filter with edge reflection or
    circular edge criterion
  - NIF: iterative version of filt1d with k-sigma clipping (Aigrain &
    Irwin 2004)

- norm: NaN-robust wrappers for min, max, mean, median, sum, plus
  routines to compute median and MAD-estimated scatter, to normalise
  to zero mean and unit variance, and to scale to range from 0 to 1.

- multiplot: routines to make plots with shared axes

- mpfit: non-linear least squares optimization using
  Levenberg-Marquart algorithm, with option to fix or bound some
  parameters. Adapted by me from a really old version that was itself
  adapted from the IDL routine of the same name. Horribly cumbersome
  and rather unstable but quite useful because very generic.

- show_cmaps.py: little script to display the available color maps

- extrema.py: returns indices and values of the extrema of an array

- GP_simple: ultra-simple GP regression code, useful to show complete
  GP newbies just how simple it all is

- GP_suz: a slightly more sophisticated GP regression code designed to
  be called from other programs. Includes a range of kernel functions,
  and a transit mean function (using Tom Evans's planetc module) - ask
  me if you want some other kernel or mean functions added. Also
  includes functions to tune the parameters using max-likelihood
  (scipy.optmize.fmin) or a simple Metropolis-Hastings MCMC. The
  likelihood can be combined with priors to explore a posterior
  distribution but at the moment only Gaussian priors are implemented.

- spot_model: generate light and radial velocity curves for active
  stars using a very simple spot model (no faculae, no
  penumbra). There are two options, one that includes limb-darkening
  and projection effects over the spots, using the formalism of Dorren
  (1989) for circular spots, but is slow, and one that ignores these
  effects, but is very fast. The user can set the stellar radius and
  inclination, the (linear) limb-darkening coefficient for the
  photosphere, the number of spots, and the effective size (size *
  contrast), longitude, latitude, limb-darkening coefficient, and
  rotation period of each spot. Spot evolution can also be included,
  in this case the effective size of each spot varies according to a
  bell-shaped (Gaussian) curve with standard deviation and peak time
  specified by the user. The code incluces user-friendly routines for
  generating and plotting either regularly sampled time-series, or
  irregularly-sampled time-series more typical of ground-based
  observations, as well as their periodograms.

- orbit: compute sky coordinates and radial velocities for star or
  planet in two-body Keplerian system.