Saeed’s Data Visualization Blog

  My visualization was produced very smoothly using ChatGPT. I asked it to produce a scatterplot of Sepal Length versus Sepal Width from the iris dataset, so the code it supplied ran almost instantly within RStudio. I liked how quickly I could alter plot colours, titles and themes - just a few minor adjustments - and it worked. My primary concern was to make certain the saved image exported at a sufficient resolution, but adding the ggsave () function solved that problem ChatGPT served as a quick assistant - it wrote the first clean draft so I could finish the design without losing control.

  I carried out the task in Python. I used NetworkX to build the graph besides Plotnine to draw it. After I installed the right Python version and the packages, the steps ran without trouble. The hardest part was the first install - Windows did not find pip or the correct Python runtime. Once I fixed that, the script ran and gave a clear network plot. Python needs more steps at the start than R - yet NetworkX or Plotnine gives ample room to adjust the picture. I would definitely use this method again for network visualizations, especially when working in a Python-based workflow.

  I rebuilt the dot - dash rug scatterplot from Lukasz Piwek's Tufte in R guide because it shows how to keep ink that carries data and drop ink that does not. The chart removes grid lines, thick borders plus repeated words - only the marks that carry facts remain. Short ticks added to both axes reveal where values stack and where they sit alone. The result is a calm, spare graph that follows Tufte's rule - show the numbers nothing extra. I opened R Studio and loaded the built in dataset mtcars. I installed ggplot2 plus ggthemes. I typed the supplied commands - geom_point() drew the scatterplot, geom_rug() added the small tick marks on the sides and theme_tufte() gave the plain Tufte look. The graph that appeared plotted weight in thousands of pounds against miles per gallon - it matched the requested minimalist style.

  The charts were drawn in R with the ggplot2 package and the built in economics file. They show the link between the share of people out of work (unemploy divided by pop) plus the median number of weeks people stay unemployed (uempmed) in the United States from 1967 through 2010. The first chart draws a line that links the two indicators for every month - the second chart gives each dot a color that matches its year. Both charts show that when the jobless share goes up, people also stay unemployed longer - the jump in duration is sharpest during recessions. A continuous line and a color scale that runs from light to dark turn decades of data into a single picture that exposes long run shifts and repeated downturns at a glance. A table of numbers cannot deliver the same immediate grasp of how the labor market contracts but also expands year after year. Reference: Yau, N. (2011). Visualize This: The FlowingData Guide to Design, Visualization, and Statistics. Wiley.