Any more Stan or `brms` functionality we can use?

[athowes]

In this package we are injecting custom Stan code to allow us to fit the models we would like to fit.

It would be preferable if we didn't do this, and instead relied on existing Stan or brms functionality where exists.

We should investigate:

• Truncation via T[]
• Censoring via cens() (?)
• Anything else in the manuals that could be helpful

[athowes]

According to https://mc-stan.org/docs/stan-users-guide/truncation-censoring.html#truncation.section:

• Truncation in Stan is restricted to univariate distributions for which the corresponding log cumulative distribution function (CDF) and log complementary cumulative distribution (CCDF)
• More reading...

We have the code:

family_lpdf(d | mu, sigma) - family_lcdf(obs_time | mu, sigma);

I guess that we could rewrite this as:

family_lpdf(d | mu, sigma) T[0, obs_time];

In some sense, this is nice. It makes it clearer what is going on. In another sense, what is the point? Do we have any way of getting out of the custom functions? Need this to be in brms.

[athowes]

brms apparently does allow for both truncation and censoring: https://github.com/paul-buerkner/brms/issues/172

[athowes]

I tried using T[] in #202 to replace some part of the text in functions.stan but found it challenging to do and dropped it.

[seabbs]

it doesn't support double event censoring. The censoring support with truncation was used in one of the models we showcase in the paper (the uniform approx)

[seabbs]

I'd suggest we close this as won't do?