Chapter 6.

1 : Report the results using data visualization

“Visualizations gives you answers to questions you didn’t know you had” – Schneiderman
Basics of data visualization:
93% of all communication is nonverbal
70% of all sensory receptors are in eye
90% of information transmitted to brain is visual
40% of people respond better to visual than other
50% of your brain is active in visual processing

 presents the results of the analysis in an accessible way using charts/graph

Charts: graph, table, and diagrams
- Components: data series (raw data – data point), Y-axis, Scale, Tick Marks (appear that
equal intervals 0 - … (first number is increment)), sort order (based on numerical –
ascending/descending), X-axis

Infographics= graph with minimal text

- For used on social media to present complicated information
- Break down difficult concepts into categories and visual component that are easy to read
and understand
Datasets similar to statistical properties but completely different data plots = need data
visualization

Tools:
Excel: add a chart to numerical analysis
Power BI & Tableau: large datasets, blend/merge datasets, charts, dashboards

6.2: Distinguishing among chart types

Selection:
1. What business question are we trying to answer

2. Who is the target audience for the visualization?

3. Which type of data is being visualized, categorical or numerical? art > science
Charts for categorical:
1. Comparison:
- Pie chart: 5 or less categories, slides is equal to proportion (# observation/ grand total) of
each category
- Bar chart: proportion of each category, easy to interpret b/c brain is more skilled at
comparing the height of columns, stacked bar charts, discrete (whole number+ finite)
- Stacked bar chart: difficult to read, confuse audience if more than 4/5 layers or too many
bars, take less space. Good when break down and compart part of the whole
- Tree map: use size and color, +: pattern, outliers, many categories. - : not exact
proportion/number (need to add labels showing number
 - Clustered bar: two categorical variables. Bares are grouped by positions for 1 categorical,
and color indicating second categorical
Heat map: +: visualize data across multiple categorical, patterns, correlation. Different
shapes of colors to indicate different values so no exact proportion or number
2. Geographic data: little dots unlike filled which is shading
- Symbol map: +: proportions across geographical areas, - : not exact proportion or number
3. Text data: word cloud: frequency of each word mentioned in a data
base, high frequency, large + bold font. Good for open ended
questions

Charts for numerical: above but word cloud works for them too since more
complex the following are best:
1. Outlier detection:
- Box and whisker plots and histogram: distribution and identify outliers

Answers why

2. Changing over time:

- Line charts: trends over time, data changes+ continuous
3. Relationship:
- Scatterplots: 2 variables on own axis, correlation between two variables
4. Geographical
- Filled geographic maps: how values differ over geographic region

4. What type of analysis have you performed exploratory or confirmatory

 Exploratory: historical data, descriptive & diagnostic analytics
- Initial investigations on historical data to summarize and explain things
- Ex: checking assumptions, finding outliers, detecting pattern
- Charts: bar, pie, line charts (sort chronologically, trend not precise)
- Describing past performances, exploring summary performance, statistics, identifying
anomalies, and outliers
 Confirmatory: probability of future event, predictive & prescriptive analytics
- After understanding the relationship in the data
- Uses statistics to judge likelihood of future event or outcome occurring

Charts for categorical data in Tableau:

- Drag desired dimension to text on marks card, drag same dimension to size on marks
card, right click on dimension on the size and select measure than count
- Automatic to text, add color by dragging, leave “the” “an” out
- Frequency highest = larger and bolder font

6.3 Visualizing exploratory business analytics:

Exploratory visualizations:
Line chart:

Histogram VS bar chart:

H: represents bins/intervals (subset of data arranged in increasing order – each bins = same sie),
use numerical data, No gaps, Y-axis is frequency/count of observations
B: represent categories, gaps, y-axis can be variety of descriptive statistics

Deceptive Charts:
- Representations intentionally or unintentionally designed to mislead the viewer about
data being presented: Manipulates scales, axes, or data to create false perceptions
Bars must always have 0 start and give wrong impression if divided into very large or very small
increments
Be careful with dual axis: arbitrarily choosing axis ranges, can make different data series look as
correlated as we like
Make sure scatter plot is standard direction
3D graphs can be deceptive b/c may distort the scale even when number are accurate

Times series data: chronologically

Bar chats = sorting based on the numerical value of the variable
Line chart = sorting chronologically + flexible with scales + can start below 0

Visualizing outliers and anomalies:

1. eyeball data to look if any come out distribution
2. Construct a box plot

Pivot tables + pivot charts: for exploratory