stan runtime error "Exception: boost::too_many_args: format-string referred..."

[Will Morris]

Hi all,

Could someone please help me out with this runtime error (below) I am getting. Have tried to run this model (http://cl.ly/131S3n330Z1K, with data http://cl.ly/code/0G3z0o441S0E) on both a Mac and Linux with optimisation on and off. Both give this error I can't interpret. Stan (v-1.1.0) itself is working fine as I can run the example models without error.

Cheers,

STAN SAMPLING COMMAND

data = HKKdata.R

init = random initialization

init tries = 1

samples = samples.csv

append_samples = 0

save_warmup = 0

seed = 3176479917 (randomly generated)

chain_id = 1 (default)

iter = 2000

warmup = 1000

thin = 1 (default)

equal_step_sizes = 0

leapfrog_steps = -1

max_treedepth = 10

epsilon = -1

epsilon_pm = 0

delta = 0.5

gamma = 0.05

Exception: boost::too_many_args: format-string referred to less arguments than were passed

Diagnostic information: 

Throw location unknown (consider using BOOST_THROW_EXCEPTION)

Dynamic exception type: boost::exception_detail::clone_impl<boost::exception_detail::error_info_injector<boost::io::too_many_args> >

std::exception::what: boost::too_many_args: format-string referred to less arguments than were passed

[Daniel Lee]

Thanks for bringing this up. I haven't looked at your model, but this error message looks like a bug in Stan. I will try to hunt it down tomorrow. 

Daniel

--
 
 

[Ben Goodrich]

On Wednesday, December 19, 2012 10:06:42 PM UTC-5, Will Morris wrote:

Could someone please help me out with this runtime error (below) I am getting. Have tried to run this model (http://cl.ly/131S3n330Z1K, with data http://cl.ly/code/0G3z0o441S0E) on both a Mac and Linux with optimisation on and off. Both give this error I can't interpret. Stan (v-1.1.0) itself is working fine as I can run the example models without error.

This is our bug. You can work around it by 1) restricting xi be slightly positive as in

real<lower=0.000001> xi[3];

and by 2) moving the calculation of the rhos into a generated quantities block at the end instead of the transformed parameters block

generated quantities {
  real<lower=-1, upper=1> rho[3];
  rho[1] <- Sigma_B_raw[1, 2] / sqrt(Sigma_B_raw[1, 1] * Sigma_B_raw[2, 2]);
  rho[2] <- Sigma_B_raw[1, 3] / sqrt(Sigma_B_raw[1, 1] * Sigma_B_raw[3, 3]);
  rho[3] <- Sigma_B_raw[2, 3] / sqrt(Sigma_B_raw[2, 2] * Sigma_B_raw[3, 3]);
}

In addition, although this is largely unrelated to the bug, if you are going to specify a uniform (0,100) prior on xi, then you should specify an upper bound on xi's support, like

real<lower=0.000001, upper=100> xi[3];

in which case, you do not need to explicitly specify the uniform prior on xi in the model block.

With all that, your model seems to run, albeit painfully slowly. There is a lot of new stuff in the Stan manual for v1.1.0 on reparameterizing models like this in Chapter 13.

Ben

[William Morris]

Thanks Ben and Daniel.

The slightly positive lower bound on `xi` did the trick. But I get the same error if I try to specify an upper bound of 100 instead of the uniform prior.

Cheers,

Will Morris M.Phil.

--- http://wkmor1.wordpress.com

Publications: Google Scholar

Twitter: @wkmor1

Skype: wkmor1

Phone: +61 3 8344 8086

----------------------------------------------------------------------

Quantitative and Applied Ecology Group

--- http://qaeco.com

The School of Botany

The University of Melbourne

----------------------------------------------------------------------

Environmental Decisions Group

--- http://www.edg.org.au/

ARC - Centre of Excellence for 

Environmental Decisions

--- http://ceed.edu.au/

NERP Environmental Decisions Hub

--- http://www.nerpdecisions.edu.au/

----------------------------------------------------------------------

$_$__ ,

 U___ }

  | |   | |

--
 
 

[Daniel Lee]

Thanks for bringing up the issue. That was definitely a bug in Stan. The fix will appear in the next release. (commit 25509af)

If you're willing, try your model with the current source. As of today, the main branch is pretty stable.

Daniel

--