Sunday, February 04, 2007
Lab 1 - Bivariate Maps
I looked at data last semester on grandparents responsible for their grandchildren; so, I continued that research by looking at whether there was a correlation with child poverty rates. I began by doing two separate maps of the grandparent phenomena and the child poverty rates. The grandparent data distribution was platykurtic with a slight negative skewing and the child poverty data distribution was mesokurtic with positive skewing; so, I chose to use the quantiles method of classification. I used a blue color scheme for the grandparent data and a warm cream to red scheme for the poverty data. In both cases, the lighter colors represented lower percentages and the dark ones higher percentages. Regional patterns were obvious as can be seen in the two maps below.
For the bivariate map, I combined the two color schemes: the poverty cream blending with the grandparent blues to yield a green gradient, the poverty red blending with the grandparent blues to yield a fushia/purple gradient, etc. This color scheme split the legend into a blue/green area of higher grandparent rates but lower poverty and a fuschia/pink area of higher poverty but lower grandparent rates with the apex in deep purple for high rates of both phenomena. The two data sets only had a .54 correlation for the country as a whole, but there were regional patterns. So, the use of this particular color scheme highlighted the patterns where they existed, particularly the upper Plains in blue/green, the Sun Belt in purple/fuschia, and the pale green North Atlantic seaboard.
One of the most difficult aspects of the mapping was finding a suitable background shade that would work with all three maps and not make Lake Michigan look too much like a state. I chose a metallic gold with a diagonal gradient which provided fairly good contrast for all three maps. Also, despite saving at the highest possible quality setting, I'm still having trouble with displaying in Blogger. The images look crisp as they're uploading but the screen then "snaps" to show the whole image at which point clarity is lost. I had problems with this last semester, too, and still haven't figured out the solution.
For the bivariate map, I combined the two color schemes: the poverty cream blending with the grandparent blues to yield a green gradient, the poverty red blending with the grandparent blues to yield a fushia/purple gradient, etc. This color scheme split the legend into a blue/green area of higher grandparent rates but lower poverty and a fuschia/pink area of higher poverty but lower grandparent rates with the apex in deep purple for high rates of both phenomena. The two data sets only had a .54 correlation for the country as a whole, but there were regional patterns. So, the use of this particular color scheme highlighted the patterns where they existed, particularly the upper Plains in blue/green, the Sun Belt in purple/fuschia, and the pale green North Atlantic seaboard.
One of the most difficult aspects of the mapping was finding a suitable background shade that would work with all three maps and not make Lake Michigan look too much like a state. I chose a metallic gold with a diagonal gradient which provided fairly good contrast for all three maps. Also, despite saving at the highest possible quality setting, I'm still having trouble with displaying in Blogger. The images look crisp as they're uploading but the screen then "snaps" to show the whole image at which point clarity is lost. I had problems with this last semester, too, and still haven't figured out the solution.# posted by Anne Dement @ 12:56 PM 
