explanation of methods

index.html

@@ -18,6 +18,7 @@
       border-collapse: collapse;
       border: 2px black solid;
       font: 12px sans-serif;
+      text-align: center;
     }
     td {
       border: 1px black solid;
@@ -72,6 +73,34 @@ <h1>2020 United States Presidential Election Model</h1>
     </p>
   </header>
 
+  <section>
+    <h2>Methods</h2>
+    <p>
+      <b>Disclaimer</b>: This model is not an official prediction. I am in no way attesting to its potential accuracy or claiming it to be correct. 
+      The point of this project was to be a learning exercise for me as well as an outlet for an interest of mine. The data you see here is purely a result 
+      of my model and I am not in any way a professional election forecaster. Nonetheless, I hope this peaks your interest.
+    </p>
+    <p>
+      When I first decided to begin working on this model in August, I wanted to challenge myself to learn more about different Python libraries, 
+      such as Scipy, Numpy, Pandas, and Matplotlib and to see if I could come up with anything accurate at all. I utilized my prior skills in webscraping 
+      and data manipulation while using new knowledge in statistics and data visualization. The entire model is written in Python.
+    </p>
+    <p>
+      This model essentially uses election data from 2016 and polling for this current election cycle to create a rough forecast of the 2020 election.
+      I gathered data about all 50 states and the District of Columbia from <a href="https://www.census.gov/">the census website</a>, 
+      2016 election data from <a href='https://www.fec.gov/'>the FEC website</a>, 2020 election polling data from <a href='https://www.270towin.com/'>270 to win</a>, 
+      and pollster ratings and bias from <a href='https://fivethirtyeight.com/'>538</a>.
+    </p>
+    <p>
+      After the data was gathered and cleaned, I had three classes of objects: States, Polls, and Pollsters. I wanted to use the data I gathered about each of these objects, 
+      to implement this model. The model thus breaks down into these parts: correcting polls based on the grade and historical bias of their pollsters, 
+      aggregating polls in each state into a singular rating with a vote share for both candidates and margin of error based on the combined margins of all the polls 
+      (for states with no polling data, this rating was simply based on the 2016 election results, as states with no polling data are not assumed to be competitive 
+      or chance much in their results from the prior election cycle), combining the prior election results and the ratings for each state into distributions of possible vote share 
+      for each candidate in each state, simulating 10,000 races in each state, and visualizing the data with the map and graphs you see below.
+    </p>
+  </section>
+
   <div class="container">
     <h2>Average Margin of Victory</h2>
     <img src="results/map.png" alt="Average Margin of Victory" width="800" height="536">