visualization for discovery
TRANSCRIPT
Jeffrey Heer @jeffrey_heer Univ. of Washington + Trifacta
Visualization for DISCOVERY
Effectiveness of Penicillin, Neomycin & Streptomycin vs. Bacteria Species
Which antibiotic should one use?
Show data variation, not design variation.
- Edward Tufte
Aerobacter aerogenes
Brucella abortus
Bacillus anthracis
Diplococcus pneumoniae
Escherichia coli
Klebsiella pneumoniaeMycobacterium tuberculosis
Proteus vulgaris
Pseudomonas aeruginosa
Salmonella typhosa
Salmonella schottmuelleri
Staphylococcus albus
Staphylococcus aureus
Streptococcus fecalis
Streptococcus hemolyticus
Streptococcus viridans
0.0010.010.11101001,000
Log10(1 / Neomycin)
0.001
0.01
0.1
1
10
100
1,000
Log1
0(1
/ Pen
icill
in)
GenusStreptococcus
Staphylococcus
Salmonella
Other
Wainer & Lysen Am. Sci. 2009
Aerobacter aerogenes
Brucella abortus
Bacillus anthracis
Diplococcus pneumoniae
Escherichia coli
Klebsiella pneumoniaeMycobacterium tuberculosis
Proteus vulgaris
Pseudomonas aeruginosa
Salmonella typhosa
Salmonella schottmuelleri
Staphylococcus albus
Staphylococcus aureus
Streptococcus fecalis
Streptococcus hemolyticus
Streptococcus viridans
0.0010.010.11101001,000
Log10(1 / Neomycin)
0.001
0.01
0.1
1
10
100
1,000
Log1
0(1
/ Pen
icill
in)
GenusStreptococcus
Staphylococcus
Salmonella
Other
Wainer & Lysen Am. Sci. 2009
What does antibiotic response reveal about the
biology of bacteria?
Aerobacter aerogenes
Brucella abortus
Bacillus anthracis
Diplococcus pneumoniae
Escherichia coli
Klebsiella pneumoniaeMycobacterium tuberculosis
Proteus vulgaris
Pseudomonas aeruginosa
Salmonella typhosa
Salmonella schottmuelleri
Staphylococcus albus
Staphylococcus aureus
Streptococcus fecalis
Streptococcus hemolyticus
Streptococcus viridans
0.0010.010.11101001,000
Log10(1 / Neomycin)
0.001
0.01
0.1
1
10
100
1,000
Log1
0(1
/ Pen
icill
in)
GenusStreptococcus
Staphylococcus
Salmonella
Other
Wainer & Lysen Am. Sci. 2009
Aerobacter aerogenes
Brucella abortus
Bacillus anthracis
Diplococcus pneumoniae
Escherichia coli
Klebsiella pneumoniaeMycobacterium tuberculosis
Proteus vulgaris
Pseudomonas aeruginosa
Salmonella typhosa
Salmonella schottmuelleri
Staphylococcus albus
Staphylococcus aureus
Streptococcus fecalis
Streptococcus hemolyticus
Streptococcus viridans
0.0010.010.11101001,000
Log10(1 / Neomycin)
0.001
0.01
0.1
1
10
100
1,000
Log1
0(1
/ Pen
icill
in)
GenusStreptococcus
Staphylococcus
Salmonella
Other
Wainer & Lysen Am. Sci. 2009
?
?
Aerobacter aerogenes
Brucella abortus
Bacillus anthracis
Diplococcus pneumoniae
Escherichia coli
Klebsiella pneumoniaeMycobacterium tuberculosis
Proteus vulgaris
Pseudomonas aeruginosa
Salmonella typhosa
Salmonella schottmuelleri
Staphylococcus albus
Staphylococcus aureus
Streptococcus fecalis
Streptococcus hemolyticus
Streptococcus viridans
0.0010.010.11101001,000
Log10(1 / Neomycin)
0.001
0.01
0.1
1
10
100
1,000
Log1
0(1
/ Pen
icill
in)
GenusStreptococcus
Staphylococcus
Salmonella
Other
Wainer & Lysen Am. Sci. 2009
Not a streptococcus!
Actually a streptococcus!
How might our tools spur new questions and
prompt skepticism?
0 10 20 30 40
Neomycin
0
200
400
600
800
Penicillin
0.001 0.01 0.1 1 10 100
Log10(Neomycin)
0.001
0.01
0.1
1
10
100
1,000Lo
g10(Pe
nicillin)
0.0010.010.1110100
Log10(1 / Neomycin)
0.001
0.01
0.1
1
10
100
1,000
Log1
0(1
/ Pen
icill
in)
0.0010.010.11101001,000
Log10(1 / Neomycin)
0.001
0.01
0.1
1
10
100
1,000
Log1
0(1
/ Pen
icill
in)
0.0010.010.11101001,000
Log10(1 / Neomycin)
0.001
0.01
0.1
1
10
100
1,000
Log1
0(1
/ Pen
icill
in)
GenusAerobacterBrucellaBacillusDiplococcusEscherichiaKlebsiellaMycobacteriumProteusPseudomonasSalmonellaStaphylococcusStreptococcus
0.0010.010.11101001,000
Log10(1 / Neomycin)
0.001
0.01
0.1
1
10
100
1,000
Log1
0(1
/ Pen
icill
in)
GenusStreptococcus
Staphylococcus
Salmonella
Other
Aerobacter aerogenes
Brucella abortus
Bacillus anthracis
Diplococcus pneumoniae
Escherichia coli
Klebsiella pneumoniaeMycobacterium tuberculosis
Proteus vulgaris
Pseudomonas aeruginosa
Salmonella typhosa
Salmonella schottmuelleri
Staphylococcus albus
Staphylococcus aureus
Streptococcus fecalis
Streptococcus hemolyticus
Streptococcus viridans
0.0010.010.11101001,000
Log10(1 / Neomycin)
0.001
0.01
0.1
1
10
100
1,000
Log1
0(1
/ Pen
icill
in)
GenusStreptococcus
Staphylococcus
Salmonella
Other
A Combinatorial Design Space
1. Variable Selection
A Combinatorial Design Space
1. Variable Selection
2. Data Transformation
A Combinatorial Design Space
1. Variable Selection
2. Data Transformation
3. Visual Encoding Design
A Combinatorial Design Space
1. Variable Selection
2. Data Transformation
3. Visual Encoding Design
—> Thousands of possible charts!
A Combinatorial Design Space
How might we augment manual chart construction
with interactive browsing of recommended visualizations?
ggplot(diamonds, aes(x=price, fill=cut))+ geom_bar(position="dodge")
1. Trifacta Visual Profiler
2. Data Voyager (UW + Tableau)
1. Trifacta Visual Profiler
2. Data Voyager (UW + Tableau)
Trifacta Visual Profiler
OverviewOverview
Trifacta Visual Profiler
DetailsDetails
Trifacta Visual Profiler
Trifacta Visual Profiler
Trifacta Visual Profiler
Trifacta Visual Profiler
Trifacta Visual Profiler
Trifacta Visual Profiler
Trifacta Visual Profiler
1. Trifacta Visual Profiler
2. Data Voyager (UW + Tableau)
User
Voyager Visualization Browser
User
Data Set
Voyager Visualization Browser
Compass Recommendation Engine
Data Schema & Statistics
User
Voyager Visualization Browser
Compass Recommendation Engine
Data Schema & Statistics
User
1. Select data variables 2. Apply transformations 3. Pick visual encodings
Voyager Visualization Browser
Compass Recommendation Engine
Data Schema & Statistics
User
Constrain & rank choices by data type, statistics & perceptual principles.
Voyager Visualization Browser
Data Schema & Statistics
Ranked and Clustered Vega-lite Specifications
User
Compass Recommendation Engine
Voyager Visualization Browser
Compass Recommendation Engine
Vega-lite Compiler
Vega-lite Specifications
Ranked and Clustered Vega-lite Specifications
User
Data Schema & Statistics
Voyager Visualization Browser
Compass Recommendation Engine
Vega Renderer
Vega-lite Compiler
Vega-lite Specifications
VegaSpecifications
Ranked and Clustered Vega-lite Specifications
User
Data Schema & Statistics
Voyager Visualization Browser
Compass Recommendation Engine
Vega Renderer
Interactive Visualizations
Vega-lite Specifications
VegaSpecifications
Ranked and Clustered Vega-lite Specifications
User
Data Schema & Statistics
Vega-lite Compiler
Voyager Visualization Browser
Interactive Visualizations
Compass Recommendation Engine
Vega Renderer
Interactive Visualizations
Vega-lite Specifications
VegaSpecifications
Ranked and Clustered Vega-lite Specifications
User
Data Schema & Statistics
Vega-lite Compiler
Voyager Visualization Browser
Interactive Visualizations
User Selection
Compass Recommendation Engine
Vega Renderer
Interactive Visualizations
Vega-lite Specifications
VegaSpecifications
User Selection, Data Schema
& Statistics
Ranked and Clustered Vega-lite Specifications
User
Vega-lite Compiler
Voyager Visualization Browser
Interactive Visualizations
User Selection
Compass Recommendation Engine
Vega Renderer
Interactive Visualizations
Vega-lite Specifications
VegaSpecifications
User Selection, Data Schema
& Statistics
Ranked and Clustered Vega-lite Specifications
User
Improves data coverage! +3x variable sets shown +1.5x more interacted with
Vega-lite Compiler
Refining visualization recommendation What to optimize? How to evaluate?
Scaling interactive visualizations Large D harder than large N…
Help avoid statistical pitfalls Recognize mix effects, convey uncertainty
Ongoing Challenges
How might our tools spur new questions and
prompt skepticism?
vega.github.io
Jeffrey Heer @jeffrey_heer Univ. of Washington + Trifacta
Visualization for DISCOVERY