Use cases: Difference between revisions

From VrlWiki
Jump to navigation Jump to search
Created page with "On this page, we list a few ways in which careful visualization designs can facilitate more statistically sound exploratory data analysis. * Using user knowledge to inform hy..."
 
No edit summary
 
Line 1: Line 1:
On this page, we list a few ways in which careful visualization designs can facilitate more statistically sound exploratory data analysis.
On this page, we list a few ways in which careful visualization designs can facilitate more statistically sound exploratory data analysis.


* Using user knowledge to inform hypothesis partition
* Visual verification of theoretically derived support threshold for frequent itemsets mining
** Assumption: users might feel uncomfortable simply ignoring itemsets below a support threshold that is derived algorithmically
** Proposed solutions: design an interface that visualizes aggregate properties of itemsets below the computed support threshold and example itemsets
* Engaging the user to reduce the space of hypotheses to be tested, e.g. include users in the hypothesis partition process
** Using user knowledge to inform hypothesis partition
*** e.g. the user might be interested only in hypotheses  that involve certain variables
** Helping the user select subsets of hypotheses to test
*** e.g. display visual summary of each set of hypotheses
* User-guided construction of null hypothesis distribution corresponding to no interesting observations
* User-guided construction of null hypothesis distribution corresponding to no interesting observations
* Understanding how each exploration action impact the overall false discovery rate
* Help the user make decisions during iterative exploratory data analysis
** e.g. choosing the appropriate alpha-investing policy

Latest revision as of 18:34, 27 February 2017

On this page, we list a few ways in which careful visualization designs can facilitate more statistically sound exploratory data analysis.

  • Visual verification of theoretically derived support threshold for frequent itemsets mining
    • Assumption: users might feel uncomfortable simply ignoring itemsets below a support threshold that is derived algorithmically
    • Proposed solutions: design an interface that visualizes aggregate properties of itemsets below the computed support threshold and example itemsets
  • Engaging the user to reduce the space of hypotheses to be tested, e.g. include users in the hypothesis partition process
    • Using user knowledge to inform hypothesis partition
      • e.g. the user might be interested only in hypotheses that involve certain variables
    • Helping the user select subsets of hypotheses to test
      • e.g. display visual summary of each set of hypotheses
  • User-guided construction of null hypothesis distribution corresponding to no interesting observations
  • Help the user make decisions during iterative exploratory data analysis
    • e.g. choosing the appropriate alpha-investing policy