control_blip

<code>control_blip</code> sets the default control arguments
for doubly robust blip-learning, <code>type = "blip"</code>.

Package for learning and evaluating (subgroup) policies via doubly robust loss functions. Policy learning methods include doubly robust blip/conditional average treatment effect learning and sequential policy tree learning. Methods for (subgroup) policy evaluation include doubly robust cross-fitting and online estimation/sequential validation. See Nordland and Holst (2022) <doi:10.48550/arXiv.2212.02335> for documentation and references.

Andreas Nordland

polle

Policy Learning

Klaus Holst

control_blip function

<dl><dt>blip_models</dt>
<dd>Single element or list of V-restricted blip-models created
by <code>q_glm()</code>, <code>q_rf()</code>, <code>q_sl()</code> or similar functions.</dd>
<dt>quantile_prob_threshold</dt>
<dd>Numeric vector. Quantile probabilities
for adaptively setting the threshold for the conditional average treatment effect.</dd></dl>

Arguments

Control arguments for doubly robust blip-learning — control_blip

<dl>

<dt>blip_models</dt>
<dd>Single element or list of V-restricted blip-models created
by <code>q_glm()</code>, <code>q_rf()</code>, <code>q_sl()</code> or similar functions.</dd>


<dt>quantile_prob_threshold</dt>
<dd>Numeric vector. Quantile probabilities
for adaptively setting the threshold for the conditional average treatment effect.</dd>

</dl>

control_blip: Control arguments for doubly robust blip-learning

Description

Usage

Value

Arguments