Description Usage Arguments Value Examples

Program for fitting a GLM equipped with a regularized student-t prior on the regression coefficients, parametrized using the normal-inverse-gamma distribution. The 'regularization' refers to the fact that the inverse-gamma scale is has a finite upper bound that it smoothly approaches. This method was not used in the simulation study but was used in the data analysis. Specifically, it corresponds to 'PedRESC2'.

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 | ```
glm_studt(
stan_fit = stanmodels$RegStudT,
y,
x_standardized,
beta_scale,
dof = 1,
slab_precision = (1/15)^2,
only_prior = F,
mc_warmup = 50,
mc_iter_after_warmup = 50,
mc_chains = 1,
mc_thin = 1,
mc_stepsize = 0.1,
mc_adapt_delta = 0.9,
mc_max_treedepth = 15,
ntries = 1,
return_as_stanfit = FALSE
)
``` |

`stan_fit` |
an R object of class stanfit, which allows the function to run without recompiling the stan code. |

`y` |
(vector) outcomes corresponding to the type of glm desired. This should match whatever datatype is expected by the stan program. |

`x_standardized` |
(matrix) matrix of numeric values with number of rows equal to the length of y and number of columns equal to p+q. It is assumed without verification that each column is standardized to whatever scale the prior expects - in Boonstra and Barbaro, all predictors are marginally generated to have mean zero and unit variance, so no standardization is conducted. In practice, all data should be standardized to have a common scale before model fitting. If regression coefficients on the natural scale are desired, they be easily obtained through unstandardizing. |

`beta_scale` |
(pos. real) constants indicating the prior scale of the student-t prior. |

`dof` |
(pos. integer) degrees of freedom for the student-t prior |

`slab_precision` |
(pos. real) the slab-part of the regularized horseshoe, this is equivalent to (1/d)^2 in the notation of Boonstra and Barbaro |

`only_prior` |
(logical) should all data be ignored, sampling only from the prior? |

`mc_warmup` |
number of MCMC warm-up iterations |

`mc_iter_after_warmup` |
number of MCMC iterations after warm-up |

`mc_chains` |
number of MCMC chains |

`mc_thin` |
every nth draw to keep |

`mc_stepsize` |
positive stepsize |

`mc_adapt_delta` |
between 0 and 1 |

`mc_max_treedepth` |
max tree depth |

`ntries` |
(pos. integer) the stan function will run up to this many times, stopping either when the number of divergent transitions* is zero or when ntries has been reached. The reported fit will be that with the fewest number of divergent iterations. |

`return_as_stanfit` |
(logical) should the function return the stanfit object asis or should a summary of stanfit be returned as a regular list |

`list`

object containing the draws and other information.

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 | ```
data(historical)
foo = glm_studt(y = historical$y_hist,
x_standardized = historical[,2:5],
beta_scale = 0.0231,
dof = 1,
slab_precision = 0.00444,
only_prior = 0,
mc_warmup = 1000,
mc_iter_after_warmup = 1000,
mc_chains = 2,
mc_thin = 1,
mc_stepsize = 0.1,
mc_adapt_delta = 0.99,
mc_max_treedepth = 15,
ntries = 2);
data(current)
foo = glm_studt(y = current$y_curr,
x_standardized = current[,2:11],
beta_scale = 0.0231,
dof = 1,
slab_precision = 0.00444,
only_prior = 0,
mc_warmup = 1000,
mc_iter_after_warmup = 1000,
mc_chains = 2,
mc_thin = 1,
mc_stepsize = 0.1,
mc_adapt_delta = 0.99,
mc_max_treedepth = 15,
ntries = 2);
``` |

Embedding an R snippet on your website

Add the following code to your website.

For more information on customizing the embed code, read Embedding Snippets.