sem_params
SEM parameters
Package |
Benchpress |
Version |
|
Language |
|
Docs |
|
Paper |
Rios et al.[1] |
Graph type |
|
Docker |
|
Module |
Description
An object of this module defines a Bayesian network by generating the weight matrix \(W\) of a Gaussian linear structural equation model (SEM) of the form
\[Y_i=\sum_{j:Y_j\in pa(Y)} W_{ij}Y_j + Z_i,\]
where \(Z_i\sim \mathcal N(\mu, \sigma^2)\) and elements of \(W\) are distributed as
\[\begin{split}W_{ij} \sim
\begin{cases}
\mathrm{Unif}([a, b])\mathrm{Unif}(\{-1,1\}) & \text{ if }(i, j) \in E\\
0 & \text{ otherwise.}
\end{cases}\end{split}\]
Example
[
{
"id": "SEM",
"min": 0.25,
"max": 1
}
]