Home

MCMCvis TODOS

0.9.0 update:

Function to preserve dimensionality of output

• specify parameters of interest and function of interest

• outputs a list - each element is a different param - some matrices, some vectors, some scalars

• stitch proper dimensionality together based on position of variable - specified as output (for instance mu[1,4])

Ex:

MCMCnewfun(coda.object, func = mean)

> list(B.matrix = round(matrix(rnorm(9, 1.5), nrow = 3), digits = 2), mu = rnorm(5), alpha = rnorm(1, 10))

MCMCnewfun(coda.object, params = 'B.matrix', func = mean)

> list(B.matrix = round(matrix(rnorm(9), nrow = 3), digits = 2))

Add instructions or examples to README on github

MCMCsummary - quantiles arg to function

• default set as quantiles = c(0.025, 0.5, 0.975) if NA, no quantiles displayed
• accept varying lengths for argument - if length of quantiles arg = 1, display only one quantiles column

MCMCsummary - scientific notation if too many 0s

• digits would be sig digits instead of digits to round to

MCMCtrace - some way to visualize posterior on top of prior

• may need to specify priors in a particular way rnorm(NUM_SAMPLE_DRAWS, PRIOR_MU, PRIOR_TAU)
• put priors in a coda object
• display degree of overlap (metric) between prior and posterior beneath plot

MCMCchains - arg to output in same dimensions as param in model with samples in Nth dim with an NxN array

• format = flat or multidimensional
• could simply grab the [1,1] format and make a cube or whatever, just as with the jags.pbject type function