Lab Report #4:  GIS & Geospatial Visualization

Introduction:

For this lab experimentation with mapping, I decided to try making small multiples of quilt patterns using Tableau. My goal is to use this software to create these individual pattern maps, and then incorporate them into the infographic that will be part of my final project.

Materials:

Tableau Public program, Paint, shape files for 1850 U.S. geography, Carto DB, Quilt Index dataset

Method:

1. I located shape files for the United States at different points in history at https://www.nhgis.org/. I loaded up CartoDB to use these in my visualization, but when I tried to upload them I was told that the file was too large, so I used Tableau.

2. After uploading my dataset to Tableau, I had to manually change the locations to states so that the program would plot the points. There were some records which had no state, so I deleted them from the working file in class, and reduced my dataset from 175 items to 94.

3. The first map I created had all of the patterns in one visualization, with colors showing pattern and sizes showing the number of quilts. The visualization uses 19 different colors. 

4. I moved on to creating small multiples, using the pattern and location fields, but for some reason the dates were not being parsed correctly after I redefined the type of data in the cell. I wanted to attempt to use shades of a color in order to show the dates.

5. Finally I changed the date field back to a ‘Dimension’ and was able to achieve a gradient.

 

 

6. At home I went back to the original data set, which includes links to the item records, and found a specific location and date for each quilt. This gave me a dataset with 171 records. If the location for where a quilt was made was not clear in the full item record (in the ‘location’, ‘provenance’, ‘quilt history’, or ‘quiltmaker address’ fields), then I entered a location based on the field with the address of the owner (usually a relative/descendant of the quiltmaker) who brought the quilt in for documentation or donation. There were ~dozen instances where I entered the location of the contributor (to the Quilt Index) or the quilt collection.

7. I uploaded the file into Tableau, and created a map visualization using the pattern field, which once again gave me horizontal maps. I used the pattern field again to create columns, which gave me a grid with square maps. The ones that I need traveled in a diagonal from the top left of the visualization to the bottom right: 

8. Using Paint, I cropped screen shots of each pattern map for small multiples.

Results/Discussion:

I pasted these in to this document without distorted the size from which they are cropped, as well as the legend, so the sizes on the legend should be appropriate to the sizes of the small multiples. Overall I am pleased with the way that they turned out. Some of the maps are dense, and in order to fit them in the same size square you lose readability (Log Cabin, Sunflower, Tulip, Reel), but I think that can be sacrificed for the overall effect. It was also too bad that I couldn’t use the original shape files from 1850 that I found, with CartoDB, but ultimately since I am using small multiples, the shapes are tiny and it probably doesn’t matter.

Another issue I left alone is the problems I am having with the date field. I am going to have to enter a day and month for all of the items, and also comb through the records removing inappropriate records. In this lab report visualization, even though I selected 1800-49 as my period in the advanced search, the results returned include quilts that are dated after 1850. There are two different ways to search dates for quilts, and not all quilts have metadata in the period field. Searching a specific date range will probably give me a completely different dataset. I stuck with this one only because of the amount of time that has been invested in cleaning it so it works with different visualization tools. I can see that it is really important to prepare for anything as far as collecting metadata and creating a dataset, there are visualization tools yet to be developed that you will want to use with it in the future.

Future Directions:

I chose shades of blue for my markers because in the final visualization that I do for my project I will find ways to highlight significant details using green and red. Otherwise the background colors for the visualization will be sepia tones and grey scale. The most important thing I will have to revisit for my project is my dataset, so I will be spending some time with Google Refine this month cleaning and finalizing the data that will inform the geospatial and time-series analysis the best. There are also alternate names for some of the quilt patterns that I discovered while doing research on their origins in Barbara Brackman’s Encyclopedia of Quilt Patterns. I need to search those pattern names in the Quilt Index, and chose what range of years I want to represent. Getting the dates right will be essential for being able to do any serious analysis using the visualization!

Update April 19, 2013: I decided to select a few of the earliest examples for each pattern. Because certain patterns occurred so infrequently, like the cotton leaf, this gave me a dataset that ranged between 1-10 items for each pattern – with no data after 1940. The three fields I used are pattern, year, and location (state). I was able to save my visualization to the web with Tableau, and embed the maps into this webpage.