<?xml version="1.0" encoding="utf-8"?><feed xmlns="http://www.w3.org/2005/Atom" ><generator uri="https://jekyllrb.com/" version="3.9.0">Jekyll</generator><link href="https://yemoray.github.io/feed.xml" rel="self" type="application/atom+xml" /><link href="https://yemoray.github.io/" rel="alternate" type="text/html" /><updated>2021-07-30T01:46:16+00:00</updated><id>https://yemoray.github.io/feed.xml</id><title type="html">Yemi Adepetun</title><subtitle>I'm currently enrolled in the applied statistics and data management graduate certificate program at NC State University.</subtitle><entry><title type="html">Reflection Blog Post</title><link href="https://yemoray.github.io/Reflection-Blog-Post/" rel="alternate" type="text/html" title="Reflection Blog Post" /><published>2021-07-29T00:00:00+00:00</published><updated>2021-07-29T00:00:00+00:00</updated><id>https://yemoray.github.io/Reflection-Blog-Post</id><content type="html" xml:base="https://yemoray.github.io/Reflection-Blog-Post/">&lt;p&gt;My thoughts about what a Data Scientist does has not changed since I wrote my initial blog posts, but I have a healthy respect and appreciation for what they do especially having done
some work in Machine Learning.&lt;/p&gt;

&lt;p&gt;R has really grown on me after seeing what it can be used for, my introduction to R Shiny has been a particular highlight. I definitely plan to use R going forward, especially in
fitting supervised learning models. I also look forward to using R shiny to present data in an interactive manner (both as a hobby and a visualization tool at work). Mastering R
now will also help in the statistics courses I plan to take after re-enrolling in the masters in statistics program next Summer/Fall.&lt;/p&gt;

&lt;p&gt;Beyond learning R to a greater extent than I expected, this course has further enhanced how I approach problem-solving- simply finding a solution to a problem is not enough, doing
so in the most effective and efficient way is even better.I look forward to increasing my knowledge base on statistics to truly unleash what R can do in both statistical analysis and Machine Learning.&lt;/p&gt;</content><author><name></name></author><summary type="html">My thoughts about what a Data Scientist does has not changed since I wrote my initial blog posts, but I have a healthy respect and appreciation for what they do especially having done some work in Machine Learning.</summary></entry><entry><title type="html">Fourth Post- Project 2- Creating data summaries, predictive models, and automated Markdown reports from the Bike Sharing Dataset</title><link href="https://yemoray.github.io/Fourth-Blog-Post/" rel="alternate" type="text/html" title="Fourth Post- Project 2- Creating data summaries, predictive models, and automated Markdown reports from the Bike Sharing Dataset" /><published>2021-07-10T00:00:00+00:00</published><updated>2021-07-10T00:00:00+00:00</updated><id>https://yemoray.github.io/Fourth%20Blog%20Post</id><content type="html" xml:base="https://yemoray.github.io/Fourth-Blog-Post/">&lt;p&gt;In this project, Jiashu Zhao and I were given the task of creating data summaries, predictive models, and automated Markdown reports from the 
&lt;a href=&quot;https://archive.ics.uci.edu/ml/datasets/Bike+Sharing+Dataset&quot;&gt;bike sharing data set&lt;/a&gt; using functions from the &lt;code class=&quot;language-plaintext highlighter-rouge&quot;&gt;readr&lt;/code&gt;,&lt;code class=&quot;language-plaintext highlighter-rouge&quot;&gt;ggplot2&lt;/code&gt;, &lt;code class=&quot;language-plaintext highlighter-rouge&quot;&gt;dplyr&lt;/code&gt;, &lt;code class=&quot;language-plaintext highlighter-rouge&quot;&gt;knitr&lt;/code&gt;,&lt;code class=&quot;language-plaintext highlighter-rouge&quot;&gt;caret&lt;/code&gt;,&lt;code class=&quot;language-plaintext highlighter-rouge&quot;&gt;tidyr&lt;/code&gt;,&lt;code class=&quot;language-plaintext highlighter-rouge&quot;&gt;lubridate&lt;/code&gt;,
&lt;code class=&quot;language-plaintext highlighter-rouge&quot;&gt;leaps&lt;/code&gt;,&lt;code class=&quot;language-plaintext highlighter-rouge&quot;&gt;regclass&lt;/code&gt;, and &lt;code class=&quot;language-plaintext highlighter-rouge&quot;&gt;MASS&lt;/code&gt; packages. The day.csv file was used throughout.&lt;/p&gt;

&lt;p&gt;My roles in this collaboration included setting up the project repo, reading in the data and splitting it into the training and testing subsets, producing some of the summary
statistics and plots, creating one linear regression and Random Forest model, and writing part of the code used to compare the four models.&lt;/p&gt;

&lt;p&gt;Overall, Jiashu and I collaborated very effectively (despite the 12 hrs difference in our local timezones!) and either one of us was able to figure out the different obstacles we 
faced during the course of the project. The most difficult part for me was figuring out how to compare the four models. We elected to use the &lt;code class=&quot;language-plaintext highlighter-rouge&quot;&gt;count&lt;/code&gt; variable as the response, but 
could not use missclassification rate as a basis for comparison because the response variable is not binary. After numerous Google searches, I proposed using the RMSE as the basis,
and it seems to work quite well. Although we were able to use RMSE to select between models, the RMSE values were generally pretty high, which makes me think our model candidates 
choices leaves a lot to be desired.&lt;/p&gt;

&lt;p&gt;Like the first project, this project further increased my familiarity with R. One big takeaway from this project was that there’s no one size fits all when it comes to the best 
predictive model. During the course of the project I perused a number of journal articles that used model comparison for better response predictions (&lt;a href=&quot;https://pubmed.ncbi.nlm.nih.gov/24012917/&quot;&gt;this article&lt;/a&gt; 
was a particular highlight) and I am already thinking of how I can utilize predictive modeling with response surface methodology at work.&lt;/p&gt;

&lt;p&gt;&lt;a href=&quot;https://github.com/yemoray/ST-558-Project_2&quot;&gt;Github repo&lt;/a&gt;&lt;br /&gt;
&lt;a href=&quot;https://yemoray.github.io/ST-558-Project_2/&quot;&gt;Github page&lt;/a&gt;&lt;/p&gt;</content><author><name></name></author><summary type="html">In this project, Jiashu Zhao and I were given the task of creating data summaries, predictive models, and automated Markdown reports from the bike sharing data set using functions from the readr,ggplot2, dplyr, knitr,caret,tidyr,lubridate, leaps,regclass, and MASS packages. The day.csv file was used throughout.</summary></entry><entry><title type="html">Third Post- Project 1- Creating a vignette about functions to read and analyze data from the NHL records API</title><link href="https://yemoray.github.io/Third-Blog-Post/" rel="alternate" type="text/html" title="Third Post- Project 1- Creating a vignette about functions to read and analyze data from the NHL records API" /><published>2021-06-20T00:00:00+00:00</published><updated>2021-06-20T00:00:00+00:00</updated><id>https://yemoray.github.io/Third%20Blog%20Post</id><content type="html" xml:base="https://yemoray.github.io/Third-Blog-Post/">&lt;p&gt;In this project, I was given the task of creating a vignette that reads data from the &lt;a href=&quot;https://gitlab.com/dword4/nhlapi/-/blob/master/records-api.md&quot;&gt;NHL records API&lt;/a&gt; 
(using functions from the &lt;code class=&quot;language-plaintext highlighter-rouge&quot;&gt;httr&lt;/code&gt; and &lt;code class=&quot;language-plaintext highlighter-rouge&quot;&gt;jsonlite&lt;/code&gt; packages) and shows some basic exploratory data analysis using tools available in the &lt;code class=&quot;language-plaintext highlighter-rouge&quot;&gt;tidyverse&lt;/code&gt; package and others.&lt;/p&gt;

&lt;p&gt;I initially created functions to contact the NHL records API for 6 endpoints using the &lt;code class=&quot;language-plaintext highlighter-rouge&quot;&gt;jsonlite&lt;/code&gt; package. These functions return data frames, with all of them using
ID (franchise ID or team ID) as arguments. However, the user has the option to use both ID and team full name to access the season records, goalie records, skater records
and franchise detail. A function was also written to access the NHL Stats API. Mapping the team names to their IDs was somewhat challenging, and took a while to figure out.
It was pretty straightforward after realizing the &lt;code class=&quot;language-plaintext highlighter-rouge&quot;&gt;which&lt;/code&gt; function could be used. I then wrote a wrapper function to access all the API endpoints.&lt;/p&gt;

&lt;p&gt;Using the wrapper function, data sets, numeriical summaries and new created variables and tablees were created. Some plots were also generated using &lt;code class=&quot;language-plaintext highlighter-rouge&quot;&gt;ggplot&lt;/code&gt;. While I don’t usually
follow the NHL, some interesting trends were observed:&lt;/p&gt;
&lt;ul&gt;
  &lt;li&gt;Centers for the Dalllas Stars tend to spend the least time in the penalty box while defenders spend the most.&lt;/li&gt;
  &lt;li&gt;Teams that perform well at home during the regular season do well at home during the playoffs as well.&lt;/li&gt;
  &lt;li&gt;Home advantage did not really count for much when the &lt;em&gt;percentage&lt;/em&gt; home wins were compared to the percentage road wins during the regular season. (The wins were calculated
as win divided by the sum of win, draws and losses).&lt;/li&gt;
  &lt;li&gt;Defenders tend to spend the longest time with the Dallas Stars, although most players spent 5 seasons or less.&lt;/li&gt;
  &lt;li&gt;Teams that are very aggressive also have the best defensive solidity and vice-versa.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;Working on the project took the better part of a week, mainly because of the time spent trying to understand the APIs and how to correcting write the functions (I’ve never written more than two nested &lt;code class=&quot;language-plaintext highlighter-rouge&quot;&gt;if&lt;/code&gt; or &lt;code class=&quot;language-plaintext highlighter-rouge&quot;&gt;if else&lt;/code&gt; statements before.) The task was also executed using the limited knowledge I have of R so far, it will be interesting  to learn better ways of writing the codes for this project as the semester progresses.&lt;/p&gt;

&lt;p&gt;Overall, this project further increased my familiarity with R. It was pretty rewarding to be able to pull data from APIs and gain some insight. It’s something I’ll definitely be
interested in doing more of whether or not my career leads me down the data scientist path.&lt;/p&gt;

&lt;p&gt;&lt;a href=&quot;https://github.com/yemoray/ST-558-Project-1&quot;&gt;Github repo&lt;/a&gt;&lt;br /&gt;
&lt;a href=&quot;https://yemoray.github.io/ST-558-Project-1/&quot;&gt;Github page&lt;/a&gt;&lt;/p&gt;</content><author><name></name></author><summary type="html">In this project, I was given the task of creating a vignette that reads data from the NHL records API (using functions from the httr and jsonlite packages) and shows some basic exploratory data analysis using tools available in the tidyverse package and others.</summary></entry><entry><title type="html">Second Post- Thoughts on the R programming language</title><link href="https://yemoray.github.io/Second-Blog-Post/" rel="alternate" type="text/html" title="Second Post- Thoughts on the R programming language" /><published>2021-06-04T00:00:00+00:00</published><updated>2021-06-04T00:00:00+00:00</updated><id>https://yemoray.github.io/Second-Blog-Post</id><content type="html" xml:base="https://yemoray.github.io/Second-Blog-Post/">&lt;p&gt;Prior to this semester, the only serious statistical programming language I learned was SAS (I don’t think Minitab counts as a programming language in the true sense). Compared to SAS, my initial impression of R is that it offers a more versatile and efficient way to write code, although this comes at the cost of a much steeper learning curve. The &lt;code class=&quot;language-plaintext highlighter-rouge&quot;&gt;DATA&lt;/code&gt; and &lt;code class=&quot;language-plaintext highlighter-rouge&quot;&gt;PROC&lt;/code&gt; steps in SAS consist of statements than are mostly easy to remember, and the high-level syntax used in SAS makes coding easier to learn. R on the other hand is an object-oriented program, and may be more difficult to get used to if the user has no prior experience with other object-oriented programs like Java and C++. I learned Visual Basic in a past life, so learning the R basics was not too bad.&lt;/p&gt;

&lt;p&gt;Both programming languages have good graphic user interfaces, but I really prefer that of R because everything is accessable at a glance- from graphs generated, tables created, to help with functions. I especially love R Markdown because of how it seamlessly knits (pardon the pun :) ) code chunks and text together to creater a customized output. SAS has its ODS output systems and macros but doesn’t quite allow the same level of customization. So far, I have found the vast array of packages in R (with each having its own syntax) to be both a blessing and a curse. It’s great to be able to have several means to solve a given problem, however, no single package can do all things, although &lt;code class=&quot;language-plaintext highlighter-rouge&quot;&gt;tidyverse&lt;/code&gt; comes close, in my opinion.  SAS on the other hand has a set number of steps, which allows for greater simplicity. For instance, SAS has &lt;code class=&quot;language-plaintext highlighter-rouge&quot;&gt;PROC GLM&lt;/code&gt;, &lt;code class=&quot;language-plaintext highlighter-rouge&quot;&gt;PROC GLIMMIX&lt;/code&gt;, &lt;code class=&quot;language-plaintext highlighter-rouge&quot;&gt;PROC MIXED&lt;/code&gt;, etc. for analysing models, but they all follow similar syntax.&lt;/p&gt;

&lt;p&gt;One thing I miss the most about SAS is that it is not case-sensitive. I’ve lost count of how much time I’ve spent trying to debug an error in R only to find out that the wrong case was used for a particular object.&lt;/p&gt;

&lt;p&gt;My view on R so far is based on the limited knowledge I currently have, when compared to SAS. I expect that my impressions on R will significantly evolve as the semester progresses.&lt;/p&gt;</content><author><name></name></author><summary type="html">Prior to this semester, the only serious statistical programming language I learned was SAS (I don’t think Minitab counts as a programming language in the true sense). Compared to SAS, my initial impression of R is that it offers a more versatile and efficient way to write code, although this comes at the cost of a much steeper learning curve. The DATA and PROC steps in SAS consist of statements than are mostly easy to remember, and the high-level syntax used in SAS makes coding easier to learn. R on the other hand is an object-oriented program, and may be more difficult to get used to if the user has no prior experience with other object-oriented programs like Java and C++. I learned Visual Basic in a past life, so learning the R basics was not too bad.</summary></entry><entry><title type="html">First Post- What does a Data Scientist do?</title><link href="https://yemoray.github.io/First-Blog-Post/" rel="alternate" type="text/html" title="First Post- What does a Data Scientist do?" /><published>2021-05-22T00:00:00+00:00</published><updated>2021-05-22T00:00:00+00:00</updated><id>https://yemoray.github.io/First-Blog-Post</id><content type="html" xml:base="https://yemoray.github.io/First-Blog-Post/">&lt;p&gt;The Science Council (https://sciencecouncil.org) defines a scientist as “someone who systematically gathers and uses research as evidence, to make hypotheses and test them, to gain and share understanding and knowledge.” By this definition, a data scientist is a scientist in every sense of the word, but with research tools that distinguishes the individual from a traditional scientist. These sets of unique research tools include the ability to gather massive amount of data, clean, transform, model and visualize the data, and tell a compelling story to help both individuals and organizations solve problems and make sound decisions.  Another distinction is that, similar to statisticians, data scientists also “get to play in everyone’s backyard”, to borrow a popular quote by John Wilder Tukey, just as long as they understand the rules of the game being played. As such, data scientists need subject matter expertise to understand the nuances and relevance of the data being worked on.&lt;/p&gt;

&lt;p&gt;Apart from having domain knowledge, the two other pillars/knowledge areas a data scientist must have are statistical/analytics, and computer programming (especially languages like Java, Python, R, SAS and SQL). These 3 foundational stones provide the data scientist with the competency needed to be able to solve business (or scientific) problems by developing models, algorithms, and providing insights from data analysis. In some regard, data scientists could be viewed as statisticians, however, both roles are not quite the same.&lt;/p&gt;

&lt;p&gt;For instance, the scale of the data that data scientists and statistician work on, and the means of data collection, differ significantly. Statisticians usually rely on designed experiments, surveys and polls for small scale, usually structured data collection. On the other hand, data scientists collect numbers, texts, audio, visuals, mobile data, etc. ,usually measured in scales that could be as great as exabytes. As a result, data scientists typically spend considerable time cleaning and transforming data prior to analysis. Statisticians are often focused on establishing a single model that best fits the data and quantifies uncertainties. Data scientists rather focus on using machine learning to create models having the best predictive accuracy possible. Both roles use processes that closely follow the Scientific Method (formulate a research goal, design the study, collect data, etc.), the means in which data scientists go about this tends to be easier to digest by many more people. Despite these differences, both roles are singularly focused on distilling knowledge from data. They  also have some knowledge of statistical concepts in common, even though statisticians tend to have a deeper understanding of theoretical aspects of statistics. Both roles also rely on computational skills and power to achieve their goals, with data scientists have considerable greater programming skill sets. Also, both roles need to be effective communicators especially to stakeholders.&lt;/p&gt;

&lt;p&gt;In my current role as a material scientist, I can relate more to being a statistician. I routinely use traditional statistical concepts like ANOVA tables, p-values,margin of error etc. to differentiate the impact of different materials/additives on performance characteristics and composition of engineered goods. However, I would like to gain some of the skill sets relevant to data scientists to be able to better predict measurements like the service life and failure modes of materials without always needing to carry out expensive and time costly experiments.&lt;/p&gt;

&lt;p&gt;References:&lt;/p&gt;

&lt;p&gt;https://medium.com/odscjournal/data-scientists-versus-statisticians-8ea146b7a47f
https://mixpanel.com/blog/2016/03/30/this-is-the-difference-between-statistics-and-data-science/
https://www.simplilearn.com/data-science-vs-data-analytics-vs-machine-learning-article
https://www.springboard.com/blog/machine-learning-engineer-vs-data-scientist/&lt;/p&gt;</content><author><name></name></author><summary type="html">The Science Council (https://sciencecouncil.org) defines a scientist as “someone who systematically gathers and uses research as evidence, to make hypotheses and test them, to gain and share understanding and knowledge.” By this definition, a data scientist is a scientist in every sense of the word, but with research tools that distinguishes the individual from a traditional scientist. These sets of unique research tools include the ability to gather massive amount of data, clean, transform, model and visualize the data, and tell a compelling story to help both individuals and organizations solve problems and make sound decisions. Another distinction is that, similar to statisticians, data scientists also “get to play in everyone’s backyard”, to borrow a popular quote by John Wilder Tukey, just as long as they understand the rules of the game being played. As such, data scientists need subject matter expertise to understand the nuances and relevance of the data being worked on.</summary></entry></feed>