bgoodri
bgoodri
Yes, although that would entail writing a bunch of kernel functions, which would better be done in stan-dev/math.
It is a bit difficult to do GPs in a way that would work well for rstanarm users in general. I think we're more likely the approximate GPs like in...
For terms like (1 | g), then the variance-covariance matrix is 1x1 and the Dirichlet prior on the variance proportion would put all of its mass on 1. At that...
If you have multiple lme4-style statements, like (1 | a) and (1 | b), then the two variance-covariance matrices are each 1x1 and independent of each other before seeing the...
You might be able to pass 2 shapes and / or scales, but the gamma distribution is used for both. On Thu, Nov 12, 2020, 7:21 AM stla wrote: >...
That was supposed to work. On Thu, Nov 12, 2020 at 9:59 AM stla wrote: > I tried something like prior_covariance = decov(shape = c(50,1), scale = > c(1,1)) but...
I would guess that it is in the same order that they appear in the formula, but I wouldn't be too surprised if lme4 changes the order. On Thu, Nov...
What rstanarm does has always been slightly wrong for the reasons outlined. I don't remember the details of why it is that way off the top of my head. I...
> If I remember correctly, the issue was that the produced assembly was too big for 32-bit Windows, which meant that CRAN could not build binaries for 32-bit Windows. The...
That sounds like a good idea On Wed, Apr 25, 2018 at 5:20 PM, dpsimpson wrote: > Summary: > > The default prior for the over-dispersion parameter of the negative...