Lab Report #5: Network Visualization
My objective in this lab was to create a network visualization that is ego-centric, but focused on Gracie Mitchell, the quilter that I have been investigating on my research blog. She documented her quilt work while being interviewed by the WPA as part of the Federal Writers Project in 1938; the interviewer was shown 30 quilts and quilt tops, with 22 distinct patterns. Her interviewer also recorded the states where she lived during her life, and the years that she was in each. The dataset that I am using is from the Quilt Index, where I searched each state that she lived in for the period of time that she was there. The results a sample dataset of quilters who were her contemporaries and the patterns which were being made around her at the time; it is a sample because the dataset represents quilts that have survived, been documented by an authority, and shared publicly and collaboratively through the Quilt Index. While Gracie Mitchell’s quilts may not have physically survived, their occurrence – or instantiation – was documented by an authority, so the goal of this ego-centric network is to create a visualization that shows the context in which Gracie Mitchell quilted and chose the 22 patterns that she executed.
- Quilt Index dataset with 110 records, including 22 records for Gracie Mitchell. The fields are state, quilter, pattern, and year.
- Cytoscape – free network visualization software
- In class, after discussing multi-nodal networks, I decided to try to make a network where all of my nodes are quilters (people), allowing the edges, node sizes, and colors to show her relationship to these other women. This meant that I required a dataset with two columns, source and target, where all nodes were of the same type. I spent some time updating the dataset, and uploaded it into Gephi to make sure that it parsed correctly, but I did not have time to finish visualization.
- Since I wanted to work with a more familiar tool, I got permission to return to Cytoscape, which I have become familiar with in the past year, and after thinking about what I want to communicate clearly in my final project, I decided to risk a return to a multi-nodal network using states and quilt patterns.
- Using Cytoscape, I uploaded the state column as the source, the quilter column as the interaction, the pattern column as the target, and the year column as an edge attribute. This loads the network as a grid.
- In the Custom Graphics Manager pane I added a Sawtooth Star icon to the list. This quilt block is one of the 22 listed by Gracie Mitchell, and appropriate for a network I plan to use in a flag design. I changed all of the nodes to Sawtooth Stars, except for the states, which I found public domain state icons for, and changed the font to Courier.
- I selected the Degree-sorted Circle Layout, so that Cytoscape would run that calculation and allow me to use degree as a node attribute.
- Next I chose to use the date attribute for continuous mapping of the edge color. This allowed me to create a gradient which identifies the period when the quilt was made. I chose to leave the edges black for quilts made during the era of slavery. Otherwise the gradient fades from dark blue (older executions of the pattern) to maroon (the patterns being executed ~1938).
- I changed the node sizes so the degree of magnitude is 10x their in-degree number, and the state node sizes are 10x the number of years that Gracie Mitchell lived in each state.
- I then employed a force directed layout, and ended up with overlapping nodes. So next I chose an edge-weighted (date) forced directed layout.
- Since there was still overlap, I went in to the settings and changed the Multiplier to calculate the repulsion force to 12, and after some more experimenting I settled on a divisor of .005 to calculate the attraction force. This spaced everything out nicely, but the icons and fonts were too small so I adjusted them, keeping the states scaled to each other, and the node size to 400x the in-degree.
- Next I changed the edge line style for Gracie Mitchell to dashed lines and the edge line style for all of the other quilters to dotted lines.
- At this point I made minute manual adjustments to the positions of a few nodes, in order to minimize the overlap from the node labels and improve readability, since the final information visualization will be a static image used in combination with other static images.
- I shared my network with a few friends to get a sense of readability, and what story it was telling, and I got a suggestion to position the states relative to each other as they are geographically. I was also concerned that the network gives the impression that all of Gracie Mitchell’s quilts were made in 1938.
- After manipulating the nodes for awhile, I settled on placing the states in the center, Gracie’s quilts in an inner circle (degree sorted), and the quilts made around her in an outer circle (sorted alphabetically). This places Gracie in the center of the network, and shows how the quilt patterns she chose fit in to the larger context of where she was living and how she was influenced by the quilters around her.
- Finally I changed the background color, font sizes, and node sizes, and changed the edge color gradient to grayscale.
Ego-centric Quilt Network
This visualization shows that Gracie Mitchell probably became familiar with certain patterns while living in Texas until she was almost 40, such as the Feathered Star, Sawtooth Star, Log Cabin, and Sunflower. She possibly learned how to do the Tree of Life pattern while she was living in Chicago for 8 years. As far as the other patterns she had with her, only the Orange Peel overlaps with work that has survived from Arkansas. In comparison to other quilters of her location and era, you can infer that her pattern selection was influenced by her experiences in other states. But the fact that very few of her designs overlap with contemporaries’ shows that she was experimental, and perhaps one of the first in her area to purchase a book of patterns. She does state that she had a book of patterns in her interview, lent to a friend and never returned (as of 1938).
To really do this visualization justice, I need to use Barbara Brackman’s encyclopedia of quilt patterns, then design and upload an icon for each quilt block pattern. Seeing the actual look of each pattern in proximity to others, sized for frequency, would be a fantastic infographic on its own. This endeavor, however, is part of a larger infographic I am putting together, so I didn’t want it to be too busy or hard to read. I will also have to work on the legend, since the star size isn’t shown and has to be scaled correctly. I will probably cut and paste something together using the Cytoscape legend and Paint or Photoshop.