loglik

<code>loglik</code> computes the log-likelihood of a state-space model of specified
<code>Q/sig^2, P1/sig^2, theta1</code>.

Inference methods for state-space models, relying on the Kalman Filter or on Viking (Variational Bayesian VarIance tracKING). See J. de Vilmarest (2022) <https://theses.hal.science/tel-03716104/>.

Joseph de Vilmarest

viking

State-Space Models Inference by Kalman or Viking

loglik function

<dl><dt>X</dt>
<dd>explanatory variables</dd>
<dt>y</dt>
<dd>time series</dd>
<dt>Qstar</dt>
<dd>the ratio <code>Q/sig^2</code></dd>
<dt>use</dt>
<dd>the availability variable</dd>
<dt>p1</dt>
<dd>coefficient for <code>P1/sig^2 = p1 I</code></dd>
<dt>train_theta1</dt>
<dd>training set for estimation of <code>theta1</code></dd>
<dt>train_Q</dt>
<dd>time steps on which the log-likelihood is computed</dd>
<dt>mode</dt>
<dd>(optional, default <code>gaussian</code>)</dd></dl>

Arguments

Log-likelihood — loglik

<dl>

<dt>X</dt>
<dd>explanatory variables</dd>


<dt>y</dt>
<dd>time series</dd>


<dt>Qstar</dt>
<dd>the ratio <code>Q/sig^2</code></dd>


<dt>use</dt>
<dd>the availability variable</dd>


<dt>p1</dt>
<dd>coefficient for <code>P1/sig^2 = p1 I</code></dd>


<dt>train_theta1</dt>
<dd>training set for estimation of <code>theta1</code></dd>


<dt>train_Q</dt>
<dd>time steps on which the log-likelihood is computed</dd>


<dt>mode</dt>
<dd>(optional, default <code>gaussian</code>)</dd>

</dl>

loglik: Log-likelihood

Description

Usage

Value

Arguments