Module 8: Correlation and ggplot2

Hi everyone!

This week we learned about correlation analysis and how to utilize the ggplot2 package to visualize correlation. A great method to visualize correlation or relationships between variables is a scatterplot with a line of best fit.

Within the mtcars dataset, I wanted to gain a better understanding of how each variable correlated against each other to find a relationship between variables I find the most interesting to visualize. I used the corrplot package to generate this initial visualization.


From this plot, we can see there is a strong negative correlation (dark red) around 0.7 for horsepower (hp) and acceleration (qsec). I thought this was unusual because a car with higher horsepower would make sense to accelerate faster since it generates more power. To explore this further, I created a scatterplot with a line of best fit.

Based on this scatterplot, it is clear that increasing horsepower relates to a lower acceleration by the car. However, this graphic explains this phenomenon is more apparent within the limited range of 100 to 200 hp and 15 to 20 second accelerations. This suggests other variables should be looked into as they may be contributing to this relationship between this interval.

Few (2009) makes several recommendations for correlation visualizations including the addition of a regression line, significance of outliers weakening correlation, and avoiding the cluttering of points. I took his suggestion and added a bright red regression line to indicate the negative relationship between horsepower for a quick understanding. I can see one major outlier towards the top right where horsepower is low but the acceleration is high. However, this point does not seem to skew the regression line heavily since most of the data points are close to the regression line. I agree with minimizing cluster in points for readability. There is no evident overlapping of points in this scatterplot, so I did not have to make any additional changes to work with Few's recommendation.

Check the code out in Github!

-Ramya's POV

Comments

Post a Comment

Popular posts from this blog

Final Project: Biodiversity in U.S National Parks

Image
Hi everyone! For my final project, I plan to explore the 2016 National Park Dataset: https://www.kaggle.com/datasets/nationalparkservice/park-biodiversity/data Problem Description: Nonnative species have been identified as a major contributor to biodiversity decline because invasive species are highly adaptive to any environment and can easily outcompete natives. This loss of native species is a particular issue for U.S National Parks that were created to be untouched and protected. According to the National Park Service, they disturb ecological processes, harm ecosystem integrity, degrade natural resources, interfere with visitor experiences in parks, and exacerbate climate change and fragmentation from land use change (NPS, 2021). In order to understand these non-natives, it is important to understand their distribution, where they have been reducing biodiversity, and potential reasons to why. Problem Objectives: Visualize the largest National Parks in size and their locations in...
Read more

Module 12: Social Network Analysis

Image
Hi everyone! This week we learned about social network analysis and how we can use ggnet2 in R to accomplish this. I absolutely love the new Netflix show "Wednesday" and the original Addams Family movies, so I decided to analyze which characters in the original movie have the most interactions with the other characters.  My hypothesis:  Wednesday will have the most interactions since she is seen as the main character. The dataset site:  https://moviegalaxies.com/movies/view/26/the-addams-family/#  I imported the json file with the nodes and edges. I set the ggnet2 parameters to vary the size of the node based on the "indegree" or amount of interactions that character has overall. Result: From the visualization, we can see Granny has the most connections (9) and Wednesday has very few connections.  I enjoyed using ggnet2 to create this social network. The biggest challenge of creating this social network was finding a dataset for the visualization and customizing...
Read more