SPDsimulationTest: Goodness of Fit test, using SPD simulation

In AdrianTimpson/ADMUR: Ancient Demographic Modelling Using Radiocarbon

Goodness of Fit test, using SPD simulation

Description

Calculates the data p-value given a model

Usage

SPDsimulationTest(data, calcurve, calrange, pars, type, inc=5, N=20000, timeseries = NULL)

Arguments

data	A dataframe of 14C dates. Requires 'age' and 'sd', and at least one of 'site' and 'phase'. Optional 'datingType' comprising '14C' and/or anything else.
calcurve	A calibration curve object. Choose from intcal20 (default), shcal20, intcal13 or shcal13.
calrange	A vector of two calendar dates BP, giving the calendar range of CalArray. Can be in either order.
pars	A single vector of one parameter combination.
type	Choose from the following currently available models. Composite models can be achieved using a vector of more than one type. For example, c('norm','power') will be a composite model, where the first two parameters are the mean and SD, the 3rd and 4th parameters determine the power distribution component, for example if modelling taphonomy.
inc	Increments to interpolate calendar years. Default = 5
N	The number of simulations to generate.
timeseries	Only if 'type' = 'timeseries', a data frame should be provided containing columns x and y that define the timeseries as year and value respectively.

Details

The returned list provides various summary statistics and timeseries of the observed and simulated data:

timeseries: a data frame containing various CIs and:

calBP: a vector of calendar years BP.

expected.sim: a vector of the expected simulation (mean average of all N simulations).

local.sd: a vector of the local (for each year) standard deviation of all N simulations.

model: a vector of the model PDF.

SPD: a vector of the observed SPD PDF, generated from data.

index: a vector of -1,0,+1 corresponding to the SPD points that are above, within or below the 95% CI of all N simulations.

pvalue: the proportion of N simulated SPDs that have more points outside the 95% CI than the observed SPD has.

observed.stat: the summary statistic for the observed data (number of points outside the 95% CI).

simulated.stat: a vector of summary statistics (number of points outside the 95% CI), one for each simulated SPD.

n.dates.all: the total number of dates in the whole data set. Trivially, the number of rows in data.

n.dates.effective: the effective number of dates within the date range. Will be non-integer since a proportion of some dates will be outside the date range.

n.phases.all: the total number of phases in the whole data set.

n.phases.effective: the effective number of phases within the date range. Will be non-integer since a proportion of some phases will be outside the date range.

n.phases.internal: an integer subset of n.phases.all that have more than 50% of their total probability mass within the date range.

The default N = 20000 can be increased if greater precision is required, however this can be very time costly.

Value

Returns a list. See details.

Examples

# trivial example showing a single date can never be rejected under a uniform model:
data <- data.frame(age=6500, sd=50, phase=1, datingType='14C')
x <- SPDsimulationTest(data, 
	calcurve=intcal20, 
	calrange=c(6000,9000), 
	pars=NULL, 
	type='uniform')
print(x$pvalue)

AdrianTimpson/ADMUR documentation built on July 2, 2024, 8:39 p.m.