AnalyseMCMC is a Fortran code that can be used to analyse and present the output of SPINspiral and lalinference_mcmc.

Input options in settings file analysemcmc.dat

Place a file called analysemcmc.dat in the directory where the MCMC output files are, and where you run analyseMCMC. If no settings file is present, the code will dump default output in HTML and png format in a new subdirectory html/ The input options in the file, with their default values, are:

Basic options

Basic options:

• THIN = 10, If >1, "thin" the output; read every thin-th line
• NBURNMAX = 500000, If >=0: override length of the burn-in phase, for all chains! This is now the ITERATION number (it becomes the line number later on). Nburn > Nchain sets Nburn = 0.1*Nchain
• NBURNFRAC = 0.5, If !=0: override length of the burn-in phase, as a fraction of the length of each chain. This overrides Nburn above
• AUTOBURNIN = -1.0, If !=0: Determine burn-in automatically as the first iteration where log(L_chain) > max(log(L_allchains)) - autoBurnin. If autoBurnin<0, use Npar/2. Overrides Nburn and NburnFrac above
• MAXCHLEN = 1000000000, Maximum chain length to read in (number of iterations, not number of lines)

• FILE = 1, Plot output to file: 0-no; screen, >0-yes; 1-png, 2-eps, 3-pdf. Give an output path for files in the parameter "outputdir" below
• COLOUR = 1, Use colours: 0-no (grey scales), 1-yes
• QUALITY = 2, "Quality" of plot, depending on purpose: 0: draft, 1: paper, 2: talk, 3: poster

• REVERSEREAD = 0, Read files reversely (anti-alphabetically), to plot coolest chain last so that it becomes better visible: 0-no, 1-yes, 2-use colours in reverse order too
• UPDATE = 0, Update screen plot every 10 seconds: 0-no, 1-yes
• MERGECHAINS = 1, Merge the data from different files into one chain: 0-no (treat separately), 1-yes (default)
• WRAPDATA = 1, Wrap the data for the parameters that are in [0,2pi]: 0-no, 1-yes (useful if the peak is around 0), 2-mirror inclination from 0-pi into 0-pi/2
• CHANGEVAR = 1, Change MCMC parameters (e.g. logd->d, kappa->theta_SL, rad->deg)
• OUTPUTDIR = ".", Save output files/plots in this directory - this may be a relative or absolute path

Print options

Select what output to print to screen and write to file:

• PRSTDOUT = 1, Print standard output to 1: screen, 2: text file
• PRPROGRESS = 2, Print general messages about the progress of the program: 0-no, 1-some, 2-more, 3-debug output
• PRRUNINFO = 1, Print run info (# iterations, seed, # detectors, SNRs, data length, etc.): 0-no, 1-only for one file (eg. if all files similar), 2-for all files
• PRCHAININFO = 1, Print chain info: 1-summary (tot # data points, # contributing chains), 2-details per chain (file name, plot colour, # iterations, burn-in, Lmax, # data points)
• PRINITIAL = 3, Print starting values for each chain: 0-no, 1-yes, 2-add injection value, 3-add Lmax value, 4-add differences (~triples number of output lines for this part)

• PRSTAT = 1, Print statistics: 0-no, 1-yes, for default probability interval, 2-yes, for all probability intervals
• PRCORR = 1, Print correlations: 0-no, 1-yes
• PRACORR = 0, Print autocorrelations: 0-no, 1-some, 2-more
• NACORR = 100, Compute nAcorr steps of autocorrelations if prAcorr>0 or plAcorr>0 (default: 100)
• PRIVAL = 1, Print interval info: 0-no, 1-for run with injected signal, 2-for run without injection, 3-both
• PRCONV = 1, Print convergence information for multiple chains to screen and chains plot: 0-no, 1- one summary line, 2-add total chain stdevs, 3-add medians, stdevs for each chain

• SAVESTATS = 1, Save statistics (statistics, correlations, intervals) to file: 0-no, 1-yes, 2-yes + copy in PS
• SAVEPDF = 0, Save the binned data for 1d and/or 2d pdfs (depending on plPDF1D and plPDF2D). This causes all 12 parameters + m1,m2 to be saved and plotted(!), which is slighty annoying
• WIKIOUTPUT = 0, Save output for the CBC wiki
• TAILOREDOUTPUT = 0, Save (ascii) output for a specific purpose, e.g. table in a paper: 0-no, tO>0: use the hardcoded block tO in the subroutine tailoredOutput() in analysemcmc_stats.f90
• HTMLOUTPUT = 0, Save HTML output - experimental, partly implemented. Useful when plotting all 2D PDFs as png

Plot selection

Choose which plots to make:

• PLOT = 1, 0: plot nothing at all, 1: plot the items selected below

• PLLOGL = 1, Plot log L chains: 0-no, 1-yes
• PLCHAIN = 1, Plot parameter chains: 0-no, 1-yes
• PLPARL = 0, Plot L vs. parameter value: 0-no, 1-yes
• PLJUMP = 0, Plot actual jump sizes: 0-no, 1-yes: lin, 2-yes: log
• PLACORR = 0, Plot autocorrelations: 0-no, 1-yes
• PLRHAT = 0, Plot Rhat vs. iteration number: 0-no, 1-yes

• PLPDF1D = 1, Plot 1d posterior distributions: 0-no, 1-yes: smoothed curve, 2-yes: actual histogram. If plot=0 and savePDF=1, this determines whether to write the pdfs to file or not.
• PLPDF2D = 2, Plot 2d posterior distributions: 0-no, 1-yes: gray + contours, 2:gray only, 3: contours only. If plot=0 and savePDF=1, this determines whether to write the pdfs to file (>0) or not (=0).
• PLOTSKY = 2, Plot 2d pdf with stars, implies plPDF2D>0: 0-no, 1-yes, 2-full sky w/o stars, 3-full sky with stars
• MAPPROJECTION = 1, Choose map projection: 1-Mollweide

• PLANIM = 0, Create movie frames

Plot options

Detailed plot settings:

• CHAINPLI = 0, Plot every chainPlI-th point in chains, logL, jump plots: chainPlI=0: autodetermine, chainPlI>0: use this chainPlI. All states in between are used for statistics, pdf generation, etc.
• SCLOGLPL = 1, Scale logL plot ranges: 0: take everything into account, including burn-in and starting values; 1: take only post-burn-in and injection values into account
• SCCHAINSPL = 1, Scale chains plot ranges: 0: take everything into account, including burn-in; 1: take only post-burn-in and injection values into account

• PLINJECT = 1, Plot injection values in the chains and pdfs: 0: no, 1: yes (all pars), 2: yes ( selected pars), 3-4: as 1-2 + print value in PDF panel
• PLSTART = 1, Plot starting values in the chains
• PLMEDIAN = 1, Plot median values in the pdfs: 1-1D PDFs, 2-2D PDFs, 3-both. 4-6: as 1-3 + write value in PDF panel
• PLRANGE = 4, Plot the probability range in the pdfs: 1-1D PDFs, 2-2D PDFs, 3-both. 4-6: as 1-3 + write value in PDF panel
• PLBURN = 1, Plot the burn-in in logL, the chains, etc.
• PLLMAX = 0, Plot the position of the max logL, in the chains and pdfs

• PRVALUES = 1, Print values (injection, median, range) in pdfs
• SMOOTH = 3, Smooth the pdfs: 0 - no, >1: smooth over smooth bins (use ~10 (3-15)?). This is 1D only for now, and can introduce artefacts on narrow peaks!
• FILLPDF = 1, Fillstyle for the pdfs (pgsfs): 1-solid, 2-outline, 3-hatched, 4-cross-hatched
• NORMPDF1D = 1, Normalise 1D pdfs: 0-no, 1-normalise surface area (default, a must for different bin sizes), 2-normalise to height, 3-normalise to sqrt(height), nice to compare par.temp. chains
• NORMPDF2D = 4, 'Normalise' 2D pdfs; greyscale value depends on bin height: 0-linearly, 1- logarithmically, 2-sqrt, 3-weigted with likelihood value, 4-2D probability intervals

• NANIMFRAMES = 1, Number of frames for the movie
• ANIMSCHEME = 3, AnimScheme (1-3): determines what panels to show in a movie frame; see source code

• NIVAL = 3, Number of probability intervals
• IVAL0 = 2, Number of the default probability interval (ival0<=Nival)
• IVALS = 0.68269, 0.9545, 0.9973, 2*0.0, Probability intervals (ivals()). Values > 0.9999 will be treated as 100%. Maximum 5 intervals

Output format

Output format for plots:

• SCRSZ = 10.8, Screen size for X11 windows (PGPlot units): MacOS: 16.4, Gentoo: 10.8
• SCRRAT = 0.57, Screen ratio for X11 windows (PGPlot units), MacBook: 0.57
• BMPXSZ = 1000, X-size for bitmap (pixels): 1000 !! Too large values give incomplete 2D PDFs somehow !!
• BMPYSZ = 700, Y-size for bitmap (pixels): 700
• PSSZ = 10.5, Size for PS/PDF (PGPlot units). Default: 10.5 __ Gives same result as without pgpap
• PSRAT = 0.742, Ratio for PS/PDF (PGPlot units). Default: 0.742
• SCFAC = 1.2, !!!Not fully implemented yet!!! Scale .png plots up by this factor, then down to the x, y size indicated above to interpolate and smoothen the plot
• UNSHARP = 10, Apply unsharp mask when creating .png plots. Default: 10.

Fonts and symbols

Fonts, symbols, et cetera:

• ORIENTATION = 1, Use portrait (1) or landscape (2) for eps/pdf; mainly useful when sending a plot to a printer
• FONTTYPE = 1, Font type used for eps/pdf: 1-simple, 2-roman, 3-italic, 4-script
• FONTSIZE1D = 1.0, Scale the font size for 1D plots (chains, 1D PDFs) with respect to default. Default: 1.0
• FONTSIZE2D = 1.0, Scale the font size for 2D plots (2D PDFs) with respect to default. Default: 1.0
• CHAINSYMBOL = 1, Plot symbol for the chains: 0-plot lines, !=0: plot symbols: eg: 1: dot (default), 2: plus, etc. -4: filled diamond, 16,17: filled square,circle 20: small open circle; -10/- 11: use a selection of open/filled symbols

Plot, bin parameters

Select parameters to plot, bin, et cetera:

• NPLPAR = 14, Number of plot parameters for 1D PDFs (and chain, jump plots, max 22). Set to -1 to plot all available parameters. This is ignored if savePDF=1.
• PLPARS = 61, 62, 22, 11, 41, 31, 32, 51, 52, 13*0, MCMC parameters to plot: 11-2:tc/t40, 21-2:d^3/ logd, 31-2:RA/dec, 41:phase, 51-4:i/psi/thJo/phJo, 61-4:Mc/eta/M1/M2, 71-3:a/th/phi for spin1, 81-3: spin2
• PANELS = 2*0, Number of for 1D plots in x,y direction: 0: autodetermine

• NBIN1D = 0, Number of bins for 1D PDFs: 0: autodetermine
• NBIN2DX = 0, Number of bins in x-direction for 2D PDFs and 2D probability ranges: 0: autodetermine (for both x and y)
• NBIN2DY = 0, Number of bins in y-direction for 2D PDFs and 2D probability ranges: 0: use Nbin2Dx, -1: use Nbin2Dx*(scr/bmp/ps)rat

• NPDF2D = 3, Number of 2D-PDF plots to make: -1: all plots (91 for 12+2 parameters), >0: read parameters from the lines below
• PDF2DPAIRS = 61, 31, 52, 47*0, 62, 32, 51, 47*0, Parameter pairs to plot a 2D PDF for. Maximum 50 pairs.

I/O units used

• 0: StdErr
• 6: StdOut

• 10: input (MCMC) file
• 16: temp file in getos
• 17: temp file in timestamp
• 19: StdOut redirection (__output.txt)

• 20: Output: __statistics.dat, __wiki.dat, __bayes.dat

• 21: bsc.dat (bright star catalogue)
• 22: bsc_const.dat (BSC constellation data)
• 23: bsc_names.dat (BSC names)
• 24: milkyway*.dat (Milky Way data)

• 30: Output: __pdf1/2d.dat
• 40: Output: tailored output

• 51: Output: HTML PDF2D matrix