Data Visualization and Analysis: Nominal, Distribution, and Correlation Charts

Nominal Comparison

compares the values of a nominal variable based on the values of a second numeric variable

Visuals for Nominal Comparison

Bar chart, column chart, dot plot, lollipop chart

Distribution

shows how the values of numeric variable are distributed, or spread

Visuals for Distribution

Histogram, box-and-whisker (boxplot) chart, violin plot

Deviation

shows how a set of actual values deviate from their reference values such as budgeted or forecasted values

Visuals for Deviation

Clustered bar and column charts, gauge, bullet chart

Ranking

order the values of a variable sequentially based on the values of a second variable

Visuals for Ranking

Defined as part of a table, bar chart, or column chart

Part-to-Whole

shows how each part compares to the whole and how the parts compare to one another

Visuals for Part-to-Whole

Pie chart, donut chart, stacked bar chart, stacked column chart, treemap chart

Correlation

indicates the degree to which two variables move in the same or opposite direction

Visuals for Correlation

Scatterplot

Time Series

shows the values of a variable at sequential points in time

Visuals for Time Series

Area chart, bar chart, column chart, line chart, sparkline chart, waterfall chart

Geospatial

Assigns numeric values to locations and encodes through coloring (shade) and size (bubble size)

Visuals for Geospatial

Maps

Composite Trends

shows how a composite structure (part-to-whole) changes over time (time series)

Visuals for Composite Trends

Stacked column chart, 100% stacked column chart, stacked area chart

Pareto Analysis

Illustrates the importance of different categories (nominal comparison) ranks them (ranking), and shows their cumulative percentage (part-to-whole)

Visuals for Pareto Analysis

Pareto chart, line and column chart

Risks

Data, Analysis, Bias

Data Risks

Completeness, Accuracy, Timeliness, Internal Controls

Completeness

Is the analysis missing relevant data?

Accuracy

Are the data used in the analysis correct?

Timeliness

Are the data used in the analysis the most recent available?

Internal Controls

Were the appropriate internal controls in place to ensure the data used were correct?

Analysis Risks

Method, Data, Purpose

Method

Was the correct method used to perform the analysis?

Data

did the analysis use the right data?

Purpose

Did the analysis answer the question?

Bias Risk

Data Biases, Preparer Biases, Evaluator Biases

Data Biases

Were all the necessary and appropriate data included in the analysis?

Preparer biases

Did the preparer have any potential biases that could have influenced the preparation of the analysis?

Evaluator biases

Does the evaluator (you) have any biases that could influence the interpretation of the results?

Confirmation Bias

prove a predetermined assumption

Selection Bias

data is selected subjectively

Within-Table Numeric Calculation

X * Y

Within-Table Text Calculation

concat

Within-Table Classification

if

Across-Table Calculation

refers to other tables to calc

Single-Column Aggregation

sum/count/average of column

Filtered Aggregation

sumif, countif, averageif

Measure Hierarchies

breaks an above measure into different categories (Revenue -> US & Canada)

How Many

how many rows: count

Transaction-Who

dimension tables that describe who is involved (internal & external agents)

Transaction-What

dimension tables that describe what is involved (resource)

Transaction-When

dimension tables that describe when it happened (date)

Who-What-When Star Schema

layout that includes who, what, when tables

Integrated Star Schema

shows how who-what-when stars are connected through resources (buy, have, sell resource)

Measure

aggregate or total that can be used in reports, Center of data analytics - can be calculated and sliced in many ways

Model Reliability

consistency and stability of a predictive model's performance over time and across different datasets

How to test Model Reliability

Review statistics to see how well model fits (R2)

Model Validity

extent to predictions made by a model accurately reflect the real-world outcomes

Construct Validity

measures what it is supposed to measure

Face Validity

makes sense in the real world

External Validity

how well predictions can be generalized to new data to different populations

Internal Validity

analysis shows a true relationship or something else is causing the result