This main run script file handles all the commands to other modules attached for analysis for the OxCal and OLE approaches. Accordingly, the module first assigns all the user-defined parameters to run the entire analysis, namely, n <- 100 – this assigns how many iterations of analysis (simulations) would be needed, nsamples <- 100 – this assigns total now many sample sizes the analysis would run with, either 10, 50 or 100 for this study. true.start <- 5000 and true.end <- 3000, these two would pass the true start and end dates from within, which the random generator would later simulate dates.
rateofadoption <- 0.1 – another crucial variable in the study; this controls the rate at which the adoption of technology is controlled within the distribution of the dates. This will be assigned to the k value of the logistic growth simulator, from which random dates are generated; the higher the number, the faster the adoption rate is. As part of this study, three values are used: 0.01, 0.05 and 0.1. A logistic growth distribution was used instead of a uniform or normal as it would best replicate a society’s onset of using a technology with a low yet non-zero start and a steady S curve upwards followed by either a stagnation or a drop. This is achieved through nibleCarbon package’s rLogisticGrowth() function. Then, the derived calendar dates are back-calibrated into radiocarbon ages through uncalibrate(). Finally, error <- 20 assigns the error rate, attributed to all the randomly generated calendar dates to mimic the calibration error margin during calibration.
calender.dates<<-replicate(n,rLogisticGrowth(a=lower_earliest,b=upper_latest,k=carrying_capacity,r=growth_rate))
radiocarbon.ages <<- uncalibrate(calender.dates)$rCRA
In brief, the script first assigns empty data frames to capture the results of all four approaches. Then, it assigns the inbuilt functions, mainrun() and genscript(), followed by a parallel computing facilitation functions runfun() and runfun1(). Consequently, it first calls the mainrun() with passed-in arguments of nsamples, true.start, true.end, rateofadoption and error to generate a list of random calendar dates (saved in object radiocarbon.ages) and the error dates (error.dates) into times the total simulation n. So, a list of 100 lists of 100 dates for n <- 100 and nsamples <- 100. This is done through the external module randomgen.R’s simdates() function, whose composition will be briefed soon. The derived dates generate their respective OxCal scripts through the genscript() function, which again utilises the custom oxcalScript.R script’s oxcalScriptGen() function.
From the derived .OXCAL files, the parallel computing functions runFun and runFun1 are called to run both the OxCal models and their results are saved in lists called out and out1. Later, these lists are derived out through for loops to be cleaned through oxcalcleaner and oxcalcleanerunif called from external modules of the same names. Then, the now accumulated results in data frames called trap_df and unif_df are written out as .CSVs were directed to be named according to the sample no. and adoption rate their analysis ran with, e.g., sample_100_rate_0.01_Trap_OxCal.CSV. Consequently, with the parallel approach now over, another for loop is initiated to loop through all the lists of dates in radiocarbon.ages to run both the methods of OLE, OLE.Median and OLE.CI, also their results being written out into .CSVs in the same manner as OxCal’s results as soon as their loop is over (Figure below).
Flowchart of the OxCal modules
