Table Lens Introduction to the Table Lens concept

Table Lens • • • Introduction to the Table Lens concept Table Lens Implementation Projected Usage Scenarios Usage Comparison with Splus Critical Analysis

Interfacing with Tables • Tables as a common representation – Regularized content – Instance vs. Value Layout – Large raw tables are uninterpretable

Table Views • Two common modes of interface – Focus: User examines as many fields as will fit on her screen (660 cells) • Sacrifices relationships for detail – Context: A generated representation of broad dataset it examined • High-level interpretation with no specifics

Table Lens • Lenses are devices which focus attention on part of a large context • Table Lens allows both Focus and Context views simultaneously – Tables are regular • Deformations are also regular • Lensing creates categories of detail

Categories of detail • Central areas of lens have highest levels of detail • Row and Column focal have less detail • Non-focal areas have sharply reduced detail but are still present

Degree of Interest • Detail categorization and Visualization based on Degree of Interest calculation • DOI Translates to cell size along 2 independent axes • Binary correspondence in each dimension

Table Lens Benefits • Table Lens can display both focus and context – Much more data can be displayed at once • 30 – 100 times basic spreadsheet • Allows simultaneous view of: – – Variable value distribution shape Inter-variable correlation Specific instance values Outlier identification

Table Lens Implementation • Interactive manipulation of focus – Atomic operations – Zoom: Enlargement of focal area – Ajust: Expansion of focal contents – Slide: Positioning of focal area • Composite manipulation – Adjust-zoom: Adds items to focus while expanding focal area

Multiple Foci • Multiple focal areas are supported • Important use modifications – Adjust corrupts display – Zoom required to be global

Graphical Cell Representation • Presentation factors – – – Value Type DOI (Region) Type Cell Size User Choices Spotlighting

Other Features • • Ascending/Descending Sorting Spotlighting Formula compilation Median, Quarter, Extents Selection

Table Lens facilitates • • Correlation of variable value curves Outlier identification/interrogation Variable nesting identification Ease of use (Simple!)

Usage Comparison • Exploratory Data Analysis (EDA) • Sensemaking – “Activities in which external representations… are interpreted into semantic content and represented in some other manner”

EDA Tasks • Batch Assessment – Determining structure of information and its irregularities • Variable modeling – Finding formulaic expression for variable values

Learning Loop • Steps: – Search for representation of regularities – Encoding information into representation – Altering representation to accommodate outliers – Use of representation for discovery

Table Lens vs Splus • Estimating utility of application approach – Required time to perform tasks • Benchmark times • Empirical times – Qualitative Considerations • Ease of use • Complexity vs Return

Correlation : Table Lens

Correlation : Splus

Time-cost for important properties of all variables • Table Lens superior for iterative analysis • Splus faster for random access

Time-cost for related variables • Table Lens superior when several clusters can be grouped and eliminated early • Splus more effective when broad dataset must be analyzed

Learning costs • Table Lens performs within significant margins as well as Splus • Table Lens is much simpler than Splus