MCMC_fit_1.1: First level wrapper function for the MCMC algorithm.
In jameshay218/mcmcJH: Generic random walk MCMC algorithm for parameter inference

Description Usage Arguments Details Value See Also

View source: R/mcmc_wrapper_2.R View source: R/mcmc_wrapper_script.R

Function to provide MCMC (Adaptive Metropolis within Gibbs) fits to all data subsets from a given data frame. No return value - all relevant outputs are saved as .csv files. Brief desription:

MCMC_fit_1.1(top_dir, all_data, output_file = NULL, file_strains, param_table,
  iterations = 1000, opt_freq = 50, thin = 50, burnin = 100,
  adaptive_period = 1, nchain = 1, popt = 0.44, group = NULL,
  lower_plot_bound = 0, upper_plot_bound = 15, LIKELIHOOD_FUNCTION = NULL,
  MODEL_FUNCTION = NULL, VERBOSE = FALSE, infection_times = NULL,
  SKIP_ZEROES = FALSE)

`top_dir`	location of the top level directory to work in
`all_data`	data frame of the data to be fit. Should be of the form "group", "strain", "variable", "value", where variable is the numeric time point and value is the titre level
`output_file`	(OPTIONAL) - should be a text file to output model fitting printouts
`param_table`	the data.frame version of the parameter table read in from the .csv file above. See `load_param_table`
`iterations`	number of iterations for the MCMC algorithm
`opt_freq`	how frequently the acceptance rate is adapted. Defaults to 50
`thin`	thinning value for the MCMC chain. Default is 1
`burnin`	the length of the burn in period. Defaults to 100
`adaptive_period`	length of the adaptive period. Defaults to 1
`nchain`	number of chains to run
`popt`	the desired acceptance rate. Defaults to 0.44
`group`	the name of the current group under investigation. Mostly just used to create a directory for the group specific results
`lower_plot_bound`	lowest value to be plotted ont he y axis for model trajectory. Defaults to 0
`upper_plot_bound`	highest value to be plotted ont he y axis for model trajectory. Defaults to 15
`LIKELIHOOD_FUNCTION`	a valid pointer to an R function which returns a single, log likelihood of the data given the current parameters
`MODEL_FUNCTION`	a valid pointer to an R function which is used to evaluate the model for the current set of parameters
`VERBOSE`	flag for additional output. Defaults to FALSE
`infection_times`	an optional vector of infection event labels and times to be plotted on the model graph
`SKIP_ZEROES`	optional flag indicating whether or not all zero value groups should be skipped. Defaults to FALSE

Firstly, sets up number of cores to be used. If only one chain, then parallelises the strains. If more than one chain, puts each chain on a different core instead. Checks for any all-zero data, and removes these from further analysis. Goes through each group in the data frame and creates a new directory for each. Gets the number of strains to be fitted and goes through each one (parallel or sequentially depending on number of cores)

returns nothing!

MCMC_fit_1.2

jameshay218/mcmcJH documentation built on May 18, 2019, 11:20 a.m.