(for testing) A non-Copula sampling function as fallback

If the sample is not suited to infer a Copula (fitCopula fails), this fuction uses much simpler rules to re-draw a new sample from an older sample with some added noise.

Usage

sampleWithNoise(X, sdf = 0.01, ...)

Arguments

X

an N×M matrix (N is the sample size), M is the number of variables (MCMC or ABC vars)

sdf

factor to increase or decrease the standard deviation of the added noise

...

passed to base::sample.int()

size

size of returned sample (passed to sample.int())

Value

a matrix with size rows and nrow(X) columns.

Details

This can be used during testing, in cases where the acceptance was very low and we have to deal with a very low quality sample This function should work like base::sample, but adds small noise. Missing values are always removed.