RStan2.11.1 - error while running GP example from manual

[Wiktor Soral]

Hi,

I've encountered a problem while trying to replicate Gaussian Process simulation from the Users Guide (p. 167). I was trying to get used to the new syntax and redo some familiar analyses the new way, and this one seems broken.

Interestingly when I change ' i==j ? 0.1 : 0.0' to old version 'if_else(i==j, 0.1, 0.0)', I'm getting warnings from the parser, but everything works fine. 

I haven't seen similar topic, so I thought I'll let you know. 

Best,

Wiktor

> sessionInfo()

R version 3.3.1 (2016-06-21)

Platform: x86_64-apple-darwin13.4.0 (64-bit)

Running under: OS X 10.11.6 (El Capitan)

locale:

[1] pl_PL.UTF-8/pl_PL.UTF-8/pl_PL.UTF-8/C/pl_PL.UTF-8/pl_PL.UTF-8

attached base packages:

[1] stats     graphics  grDevices utils     datasets  methods   base     

other attached packages:

[1] rstan_2.11.1       StanHeaders_2.11.0 ggplot2_2.1.0     

loaded via a namespace (and not attached):

 [1] Rcpp_0.12.6      codetools_0.2-14 grid_3.3.1       plyr_1.8.4      

 [5] gtable_0.2.0     stats4_3.3.1     scales_0.4.0     labeling_0.3    

 [9] tools_3.3.1      munsell_0.4.3    parallel_3.3.1   rsconnect_0.4.3 

[13] inline_0.3.14    colorspace_1.2-6 gridExtra_2.2.1 

simulate_gp = "

data {

  int<lower=1> N;

  real x[N];

}

transformed data {

  vector[N] mu;

  cov_matrix[N] Sigma;

  for (i in 1:N) 

    mu[i] = 0;

  for (i in 1:N) 

    for (j in 1:N)

      Sigma[i,j] = exp(-pow(x[i] - x[j],2)) 

                  + i==j ? 0.1 : 0.0;

}

parameters {

  vector[N] y;

}

model {

  y ~ multi_normal(mu,Sigma);

}

"

simulate_gp_data = list("N" = 20,

                        "x" = seq(0,1,le=20))

simulate_gp_fit = stan(model_code = simulate_gp,

                       data = simulate_gp_data)

Error : model53762d492ad3_bc5595b9f9828bdf9502aead3874622e_namespace::model53762d492ad3_bc5595b9f9828bdf9502aead3874622e: Sigma is not positive definite:

0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

failed to create the sampler; sampling not done

[Bob Carpenter]

Thanks for reporting.

Can you send the whole example program?

Do you get the same result if you put parens
around it, as in the following?

(i==j ? 0.1 : 0.0)

The ternary operator binds very weakly, so it may just
be a bug in the manual code.

- Bob

> --
> You received this message because you are subscribed to the Google Groups "Stan users mailing list" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to stan-users+...@googlegroups.com.
> To post to this group, send email to stan-...@googlegroups.com.
> For more options, visit https://groups.google.com/d/optout.

[Wiktor Soral]

Including parens solved the issue - then it is perhaps error in the manual. Anyway thanks!

I've included the faulty example at the bottom of the previous post. Now the example look like this:

simulate_gp = "

data {

  int<lower=1> N;

  real x[N];

}

transformed data {

  vector[N] mu;

  cov_matrix[N] Sigma;

  for (i in 1:N) 

    mu[i] = 0;

  for (i in 1:N) 

    for (j in 1:N)

      Sigma[i,j] = exp(-pow(x[i] - x[j],2)) 

                  + (i==j ? 0.1 : 0.0);

}

parameters {

  vector[N] y;

}

model {

  y ~ multi_normal(mu,Sigma);

}

"

Best,

Wiktor

[Bob Carpenter]

Thanks much for pointing this out. I'll fix this and
acknowledge your help in the next manual:

https://github.com/stan-dev/stan/issues/1868#issuecomment-236656510

- Bob

> On Jul 31, 2016, at 3:44 PM, Wiktor Soral <wiktor...@gmail.com> wrote:
>