Exploring Relationships Between Variables

 For this assignment, I explored relationships between variables in the built-in mtcars dataset using a correlation analysis. The goal was to identify which car characteristics are most related to fuel efficiency and to apply Stephen Few’s visualization design recommendations for clarity and effectiveness.

 

Visualization

The heatmap above visualizes the correlation matrix for all numerical variables in the mtcars dataset. The color scale ranges from deep blue (strong negative correlation) to red (strong positive correlation), with white representing weak or no relationship.

Patterns and Relationships Observed

From the correlation matrix, several interesting relationships appear:

  • Weight (wt) and miles per gallon (mpg) have a strong negative correlation, indicating that heavier cars are less fuel-efficient.

  • Horsepower (hp) and displacement (disp) show a strong positive correlation, meaning that larger engines generally produce more power.

  • Variables associated with engine performance—such as cyl, disp, hp, and wt—tend to move together and inversely relate to mpg.

  • Transmission type (am) and quarter-mile time (qsec) show moderate relationships with other variables but not as strong as weight or horsepower.

These findings align with what we would expect in real-world automotive data: cars that are bigger and more powerful often trade off fuel efficiency for performance.

Impact of Grid Layout on Interpretation

Using a grid layout for the heatmap made it easy to compare every variable combination simultaneously. This layout reveals clusters of related variables without requiring multiple plots or scrolling through separate charts. The consistent color mapping also helped me immediately identify which relationships were strongest and in what direction, reducing cognitive load when interpreting the results.

Applying Few’s Recommendations

Stephen Few’s design principles influenced many of my choices in this visualization. I minimized visual clutter by using a neutral color palette, clear labeling, and a simple title. The absence of unnecessary borders or 3D effects ensures that the viewer’s attention stays on the data itself rather than on decorative elements.
Few’s emphasis on simplicity and direct comparison also guided my decision to use a grid structure rather than individual scatter plots. This makes the visual more analytical and easier to interpret, particularly for quickly spotting trends across multiple relationships.

Conclusion

Overall, this exercise helped me understand how correlation analysis can uncover deeper insights into variable relationships and how thoughtful design improves comprehension. The mtcars dataset provided a clear example of how visual analytics can make statistical relationships immediately visible. Following Few’s principles not only made the visualization look professional but also enhanced its communicative value by focusing purely on the data and its meaning.

Comments

Post a Comment

Popular posts from this blog

Student Loan Program Trends in the United States (2007–2025)

Image
Written Report      This project examines how federal student loan programs in the United States have changed over time, with a focus on borrowing amounts, the number of student recipients, and the growing cost burden on individual borrowers. Although student debt is often discussed as a single issue, most people are not aware that federal loans come from different programs, each with its own history and trends. The goal of this analysis is to understand which programs are expanding, which are disappearing, and whether students today are receiving more borrowed dollars per person than they did in the past.      The idea for this project was influenced by several existing visual resources, including the Federal Student Aid dashboard, DataUSA’s educational finance reports, and journalist-designed college affordability graphics. These examples revealed how powerful visual storytelling can be in making complex financial data easier to understand. However, man...
Read more

My Geographic Map of Florida COVID-19 Cases by ZIP Code

Image
     For this week’s assignment, I was instructed to find a dataset and create a visual using Tableau. I selected the Florida COVID-19 Cases by ZIP Code (April 28, 2021) dataset from the University of South Florida Libraries’ Digital Heritage and Humanities Collection, originally downloaded from the Florida Department of Health ArcGIS Data Portal. Using this data, I wanted to examine how COVID-19 cases were distributed across Florida ZIP codes during the s pring of 2021. After importing the dataset into Tableau, I generated this geographical map using a blue color scale , since it provided the clearest contrast.      On the map, ZIP codes shaded in darker blue represent higher case counts, while lighter areas represent lower counts. Unsurprisingly, South Florida ZIP codes (especially in Miami-Dade and Broward counties) had the highest reported numbers. Central Florida areas around Orlando also showed clusters of elevated case counts. What stood out most w...
Read more

Spotting Differences in R: Comparing Fuel Efficiency by Cylinder Count

Image
  For this assignment, I used RStudio and the built-in mtcars dataset to explore differences in average miles per gallon (MPG) across cars with different numbers of cylinders. Using ggplot2 , I created a simple bar chart that displays the mean MPG for 4-, 6-, and 8-cylinder cars. The chart clearly shows that 4-cylinder cars achieve the highest average fuel efficiency , followed by 6-cylinder, and then 8-cylinder vehicles with the lowest MPG. This visualization directly supports a spotting differences task, it makes the contrasts between groups immediately visible through bar height comparisons along a shared baseline. From a design perspective, I applied a minimal theme and used a clear title and axis labels to keep attention on the comparison itself. This aligns with Stephen Few’s  principle of simplifying visuals so that viewers can detect differences quickly and accurately. Likewise, Yau  emphasizes using straightforward encodings, like length along a common scale...
Read more