dataplau-blog

[dataplau-blog]

Skillspace Explorer. Explore occupations with related skills in an intuitive path-like representation of multi-dimensional "Skillspace." Clicking on an occupation will highlight its ten nearest occupations, based on 218 skill, knowledge, and work activity measurements from the [O*Net](http://onetonline.org) database. Click on one of these "neighbors" to bring up its neighbors in turn. A "breadcrumb trail" shows your past selections; clicking on a breadcrumb will take you back to that occupation. Created in D3.js, using underlying data analysis from R.

[dataplau-blog]

The heartbeat of unemployment in Wisconsin. A new visualization, built from the ground up using D3.js. At higher speeds, the seasonal patterns of unemployment in Wisconsin are akin to a heartbeat -- until the recession hits in 2008.

[dataplau-blog]

Skillspace, animated in D3.js. An animated expansion of my earlier, R-generated graphic displaying skills measures of ~800 occupations on five summary axes. Visualization produced entirely in D3.js

[dataplau-blog]

Multi-dimensional gauges, recreated in d3.js. Code and live version on bl.ocks.

[dataplau-blog]

Bubble chart, re-created using d3.js. My first d3.js visualization, after one day of learning. Code and full svg version available on Bl.ocks.org and Gist/GitHub.

[dataplau-blog]

Summarizing high-dimensional data. 218 measurements of job skills associated with >800 occupations are summarized by a Cluster Analysis and a Principal Components Analysis. Fifteen occupational clusters are summarized by their positions on the first five Principal Components (Physical/Intellectual work; Interpersonal/Technical interactions; Abstract/Concrete thought processes; Information-/Interaction- centered; Management/Production orientation). Click for full figure. Produced in R; polished in Adobe Illustrator (primarily to introduce the underlying gradient and optimize the text).

[dataplau-blog]

Depicting high-dimensional clusters in Principal Component space. >800 occupations (small points) were placed into clusters ("Groups", large points) and subclusters ("Subgroups", medium points) based on similarity across 218 skills dimensions. Principal Components Analysis was used as a data reduction technique to minimize explanatory variables. These figures show the first two Principal Components. Produced in R and Adobe Illustrator.

[dataplau-blog]

"Bubble table" demo. Employment ranges (dummy data) by occupation cluster and industry. Produced in R; cleaned in Adobe Illustrator.

[dataplau-blog]

Teaching slides: speciation mechanisms. Accumulation of genetic & phenotypic differences are depicted by changing colors. Allopatric speciation refers to speciation in regions separated by space or barriers. Parapatric speciation refers to speciation in a population continuously distributed in space, but usually experiencing an environmental gradient. Sympatric speciation refers to speciation in a population sharing the same space at some scale, though often adapting to two different microenvironments. Produced in Mathematica, Adobe Illustrator, and Keynote

[dataplau-blog]

Teaching slide: How genetic associations evolve. Demonstration of how natural selection can cause associations among genes that ultimately slow down further responses to natural selection. Produced in Mathematica, Adobe Illustrator, and Keynote.

[dataplau-blog]

Teaching slide: stable and unstable evolutionary equilibria (heterozygote advantage & disadvantage). In diploid organisms, when the fitness of individuals carrying two different versions of the same gene (heterozygotes) isn't intermediate between those carrying two copies of each of the two versions (homozygotes), there is an equilibrium point at which both versions will be present. This equilibrium point is stable if the heterozygote is the fittest genotype, but unstable if it's the least fit. This animation provides an explanation. Produced in Mathematica, Adobe Illustrator, and Keynote.

[dataplau-blog]

Cubic splines fit to simulated data, with confidence blobs demarking optima. Mean fitness of host populations coevolving with parasites as a function of parasite virulence (V) and host recombination rate. Fitness surfaces were fitted with cubic splines, with 95% confidence blobs for maxima determined by bootstrapping simulated results. Produced in Mathematica; mildly massaged in Adobe Illustrator.

[dataplau-blog]

Linked evolutionary change over time, alternate version. Static version of the "sprinkler plot", showing only the host and using colored shading to more clearly highlight deviations from the plane. Blue disk shows most recent generation; line traces trajectory over several past generations with wedges emphasizing discrete time points. Produced in Mathematica.

[dataplau-blog]

Linked evolutionary change over time ("Sprinkler plots"). This animation shows a host and a parasite species "co-evolving", with the parasite (red) "tracking" the host (blue) in genotype space. Vectors point to the current genotype frequencies; trailing dots show preceding ten generations. Genotypes in the plane are those expected if two genes (A and B) are assorting independently; vertical deviations (shown by distance between a colored vector and its black "shadow" represent non-indpendent associations between genes ("linkage disequilibrium"). Panels represent different degrees of host recombination (rho) and parasite virulence (V) Produced in Mathematica.

[dataplau-blog]

Early (~1997) 3D scatterplot with fitted surface. Demo of "lollipop graph" showing deviations of (simulated) data from fitted surface. Produced in Mathematica.