Built site for gh-pages

.nojekyll

@@ -1 +1 @@
-31b993ad
+310a276c

qmd/forecasting-ensembling.html

@@ -2073,6 +2073,10 @@ <h1 class="title"><span id="sec-fcast-ensemb" class="quarto-section-identifier">
 <section id="sec-fcast-ensemb-misc" class="level2 unnumbered">
 <h2 class="unnumbered anchored" data-anchor-id="sec-fcast-ensemb-misc">Misc</h2>
 <ul>
+<li>Papers
+<ul>
+<li><a href="https://arxiv.org/abs/2412.08916">Measuring individual model importance based on contribution to ensemble accuracy</a></li>
+</ul></li>
 <li>Statistical ensemble nearly as good as a DL ensemble and was much faster and cheaper. (<a href="https://twitter.com/MergenthalerMax/status/1598397618572259328">Thread</a>)
 <ul>
 <li>4 models used in the ensemble

qmd/geospatial-general.html

@@ -2126,6 +2126,7 @@ <h2 class="unnumbered anchored" data-anchor-id="sec-geo-gen-misc">Misc</h2>
 <li><a href="https://rspatial.org/index.html">Spatial Data Science with R and “terra”</a></li>
 <li><a href="https://ruettenauer.github.io/Geodata_Spatial_Regression/">Geodata &amp; Spatial Regression</a></li>
 <li><a href="https://rspatialdata.github.io/">rspatialdata</a>: A collection of data sources and tutorials on visualising spatial data using R</li>
+<li><a href="https://gdsl-ul.github.io/san/">Spatial Modelling for Data Scientists</a></li>
 </ul></li>
 <li>QGIS - free and open source</li>
 <li>ArcGIS - expensive and industry-standard</li>

qmd/glossary-ds-terms.html

@@ -2223,7 +2223,7 @@ <h1 class="title"><span id="sec-ds-gloss" class="quarto-section-identifier">Glos
 <li><p><span style="color: #009499"><strong>stdout</strong></span> - standard output, which is the terminal by default</p></li>
 <li><p><span style="color: #009499"><strong>TDD</strong></span> - Test-driven development is a style of programming where coding, testing, and design are tightly interwoven</p></li>
 <li><p><span style="color: #009499"><strong>TF-IDF</strong></span>- stands for term frequency-inverse document frequency, and is often used in information retrieval and text mining.</p></li>
-<li><p><span style="color: #009499"><strong>Throughput</strong></span> - the amount of material or items passing through a system or process.</p></li>
+<li><p><span style="color: #009499"><strong>Throughput</strong></span> - The amount of material or items passing through a system or process, or the number of work items finished per unit of time.</p></li>
 <li><p><span style="color: #009499"><strong>tx</strong></span> - treatment, seen as variable with different treatments as values</p></li>
 <li><p><span style="color: #009499"><strong>URI</strong></span> - Uniform Resource Identifier - a string of characters that unambiguously identifies a particular resource. e.g.&nbsp;s3//bucket/path/to/folder or http://127.0.0.1:5000or c:\Users\me\path\to\folder</p></li>
 <li><p><span style="color: #009499"><strong>UTM</strong></span> - Urchin Traffic Monitor - used to identify marketing channels</p>

qmd/stochastic-processes.html

@@ -2086,6 +2086,7 @@ <h2 id="toc-title">Table of contents</h2>
   <li><a href="#sec-stoch-hmm-msdr" id="toc-sec-stoch-hmm-msdr" class="nav-link" data-scroll-target="#sec-stoch-hmm-msdr">Markov Switching Dynamic Regression (MSDR)</a></li>
   <li><a href="#sec-stoch-hmm-msar" id="toc-sec-stoch-hmm-msar" class="nav-link" data-scroll-target="#sec-stoch-hmm-msar">Markov Switching Auto Regression (MSAR)</a></li>
   </ul></li>
+  <li><a href="#monte-carlo-simulation" id="toc-monte-carlo-simulation" class="nav-link" data-scroll-target="#monte-carlo-simulation">Monte Carlo Simulation</a></li>
   </ul>
 </nav>
     </div>
@@ -2670,9 +2671,50 @@ <h3 class="unnumbered anchored" data-anchor-id="sec-stoch-hmm-msar">Markov Switc
 <li>Adds a fraction of the residual from the previous step to the MSDR model</li>
 </ul></li>
 </ul>
+</section>
+</section>
+<section id="monte-carlo-simulation" class="level2">
+<h2 class="anchored" data-anchor-id="monte-carlo-simulation">Monte Carlo Simulation</h2>
+<ul>
+<li>A broad class of computational algorithms that rely on repeated random sampling to obtain numerical results. (<a href="https://en.wikipedia.org/wiki/Monte_Carlo_method">wiki</a>)</li>
+<li><span class="ribbon-highlight">Example</span>: Forecast Manufactured Total Output for December
+<ul>
+<li>Using a previous December’s output, you simulate how many total items will likely be produced this December.</li>
+<li>Previous December’s Daily Throughput<br>
+<a href="_resources/Stochastic_Processes.resources/mc-ex1-daily-throughput-1.webp" class="lightbox" data-gallery="quarto-lightbox-gallery-3"><img src="_resources/Stochastic_Processes.resources/mc-ex1-daily-throughput-1.webp" class="img-fluid" width="582"></a>
+<ul>
+<li>“Throughput” is the number of items produced</li>
+<li>The throughput on December 4<sup>th</sup> was 4 items</li>
+</ul></li>
+<li>Process
+<ol type="1">
+<li>Randomly sample a day in December (1-31)</li>
+<li>On the selected day, record the previous December’s throughput on that day.</li>
+<li>Repeat 31 times</li>
+<li>Sum the 31 recorded throughput values to get the first simulated total throughput for December</li>
+<li>Record the simulated total throughput value</li>
+<li>Repeat around 10K times. to get a distribution of total throughput values for December<br>
+<a href="_resources/Stochastic_Processes.resources/mc-ex1-sim-throughput-1.png" class="lightbox" data-gallery="quarto-lightbox-gallery-4"><img src="_resources/Stochastic_Processes.resources/mc-ex1-sim-throughput-1.png" class="img-fluid" width="582"></a>
+<ul>
+<li>This example sampled 100K times</li>
+</ul></li>
+<li>The mean/median of this distribution is your point forecast and quantiles are the CI.</li>
+</ol></li>
+</ul></li>
+<li><span class="ribbon-highlight">Example</span>: What’s the probability of producing a certain amound of items or more?
+<ul>
+<li>Create an ecdf using the simulated distribution counts</li>
+<li>Manual Process<br>
+<a href="_resources/Stochastic_Processes.resources/mc-ex1-ecdf-1.png" class="lightbox" data-gallery="quarto-lightbox-gallery-5"><img src="_resources/Stochastic_Processes.resources/mc-ex1-ecdf-1.png" class="img-fluid" width="265"></a>
+<ul>
+<li>Order distribution data by simulated total throughput (largest to smallest)</li>
+<li>Create a probability column by taking the cumsum of the counts and dividing by the total counts, which in this case is 100K since there were 100K simulations.</li>
+<li>There is a 1.134% percent chance that the total throughput for this December will be 47 items or greater. (i.e.&nbsp;this around the 99th quantile)</li>
+</ul></li>
+</ul></li>
+</ul>
 
 
-</section>
 </section>
 
 </main> <!-- /main -->