~Implement matrix priors for exposures (kappa, GxS) and signatures (alpha, SxC), both filled with 1s by default.~
~Throw a warning if the user selects a discrete model but provides real-valued counts, or provides integer counts for the normal model. (Original counts will be stored in counts_real, and rounded counts in counts_int, regardless of the model; counts_real will always be used for reconstruction and plotting.)~
~When opportunities are provided (including predefined options like "human-genome"), normalise opportunities matrix so that the whole matrix (not each row) adds up to G (to avoid having infinitesimal opportunities when using many samples).~
~Avoid running plot_gof and throw a warning if the range of values of N includes <4 values (the elbow method is useless for 2–3 values of N).~
~Add methylation data as test dataset.~
~Revise documentation for all functions.~
Vignette:
~Add a list of available test datasets and how to load them.~
~Emphasise Stan options iter, warmup, chains and seed, and suggest generally appropriate values.~
~For extraction with a range of values of N, explain that the minimum and maximum values should be beyond the "reasonable" range for N, and include at least four values of N, for the elbow method to work.~
~In the extraction example, add a paragraph where extracted signatures are compared against COSMIC signatures using match_signatures.~
~Add small sections at the end on the use of the normal and negbin models, mutational opportunities, signature and exposure priors, and the Dirichlet process prior.~
~Plotting arbitrary catalogues and signatures.~
~Multichain extractions using initial parameters.~
Model functions:
kappa, GxS) and signatures (alpha, SxC), both filled with 1s by default.~counts_real, and rounded counts incounts_int, regardless of the model;counts_realwill always be used for reconstruction and plotting.)~"human-genome"), normalise opportunities matrix so that the whole matrix (not each row) adds up to G (to avoid having infinitesimal opportunities when using many samples).~plot_gofand throw a warning if the range of values of N includes <4 values (the elbow method is useless for 2–3 values of N).~Vignette:
iter,warmup,chainsandseed, and suggest generally appropriate values.~match_signatures.~