syllabus‎ > ‎R: Intro‎ > ‎

Gina Castignani

Intro Chapter
What will happen if raw data is poorly transformed into derived dimensions? In what way does it affect the quality of the representation and to what extent can this be misused? In other words, when you see visual representations of to what extent are they accurate as in many cases there is a loss of intensity due to reduction? How are visualizations verifiable by the end user. I feel that an issue with reduction, particularly aggregation, in visualization is that when "massaging" and aggregating the data it is easy to manipulate its true meaning (either intentionally or unintentionally). Think Global warming and opposing theories.
Also, to what extent does the choice of the mode of visualization / type of encoding used introduce bias and is this bias significant? (For example use of colour/dotted graph on a cartesian plane where it is inappropriate due to the number of dimensions that are required to be visualized resulting in clutter and confusion or perhaps allocating certain colours to data dimensions which will make them less/more visible)

Polaris Article
What visualization aspects make this tool well suited to exploratory data analysis?
    > undo/redo - promotes exploration
    > use of color? (dependent on the end-user?)

    > In explorarory analysis one is usually seeking categorization. 3 provided by the tool. these include ord-ord, quan-ord, quan-qual.

    > emphasis on interactivity and flexibility, can easily change data/dimension ect.

    > Allows for rapid generation, focus on analysis rather than generating the visual. "Throughout the analysis both what the data users want to see and how they want to see it          changes continuously" (p. 12) Looking for general trends and patterns vs. causal relationships, requiring one to 'dig deeper' into the data and therefore different types of encoding. 

In what way does this solution uniquely facilitate the "multidimensional" aspect of the database? How would the visualization this tool provides perform in cases where the database in not relational? Would this reduce effectiveness.

Finally, although table-based may be most logical given the relational database structure is it the only/best medium of representation/ vis/encoding data? Should other forms not be included for holistic purposes?