calc_mse: Calculates the MSE over all non-zero pixels
In RaphaelRe/SOIR: Scalar-on-Image regression within Sarim

Description Usage Arguments Details Value

View source: R/calc_mse.R

Calulation is done by automatically indication non-zero pixels. See Details.

1 2	calc_mse(res, alpha = 0.05, trueBeta, info = TRUE, tol = 1e-04, reduce = mean)

`res`	matrix, contains in each row the simulated Markov chain from sarim. (e.g. by using get_beta(reduce = FALSE)) Therefore each row corresponds to a specific beta.
`alpha`	numeric between 0 and 1, defining the width of the credibility interval (see details)
`trueBeta`	the true coefficient image (e.g. smoothBeta()).
`info`	logical, if true the number of non-zero pixels will be printed.
`tol`	numeric, threshold at which value a pixel is assumed to be zero.

The MSE is calculated over the union of all non-zero pixels of the estimated and the real coefficient image. Firstly the credible inverval for each markov chain is calculated (i.e. alpha/2 for lower and upper bound). Afterwards it will be examined, if the interval contains the zero. Then all coefficients, which are non-zero (indicated through the credible interval or in the real coefficient image) will be used for the calculation of the MSE between the true and the estimated image where one has to specify the point estimatior for the posterior (default is mean).

Returns MSE

RaphaelRe/SOIR documentation built on May 28, 2019, 12:19 p.m.