Data Modeling, Analytics, and Visualization: Key Concepts and Techniques

What is a calculated column?

A new column created in the data model using a formula based on other columns in the same table; it is stored and static.

What is a measure?

A dynamically calculated total used for analysis and not stored in the data model.

When should a measure be used instead of a calculated column?

When the result depends on aggregations or must be recalculated dynamically.

What does a fact table contain?

Numeric measures and values used for analysis.

What do dimension tables contain?

Descriptive characteristics of transactions such as who, what, and when.

What is the objective of an information model?

To create rich measures and strong dimensions for slicing analysis.

What is information modeling?

Creating calculated columns, measures, and dimensions that add meaning for analysis.

What is Pattern 1 of information modeling?

Within-table numeric calculation creating a calculated column.

What is Pattern 5 of information modeling?

Single-column aggregation using functions like SUM, COUNT, and AVERAGE.

What is filtered aggregation?

Applying aggregate functions to filtered subsets of data such as using SUMIF or CALCULATE.

What is nominal comparison?

Comparing categories using a single measure.

What chart is used for distribution patterns?

A box-and-whisker plot.

What type of chart is used for correlation?

A scatterplot.

What is a ranking pattern?

Ordering categories from highest to lowest.

What is a time-series pattern?

A visualization of data over time, usually with a line chart.

What is Pareto analysis?

Identifying the "vital few" contributors, often using a line and column chart.

What are the two steps of interpreting data analysis?

Determine if the analysis makes sense, then confirm validity and reliability.

What is completeness risk?

Missing relevant data in an analysis.

What is confirmation bias?

Seeking information that supports existing beliefs.

What is the WYSIATI bias?

"What You See Is All There Is," failing to consider missing information.

What do descriptive analytics do?

Summarize categories, averages, and distributions of data.

What do diagnostic analytics do?

Explain why something occurred, such as through outlier analysis.

How can outliers be detected?

By comparing mean and median or visually plotting the data.

What is predictive analytics used for?

Predicting future outcomes, often using regression.

What does adjusted R-squared measure?

How well the regression model fits the data.

What does standard error measure?

Accuracy of predicted values compared to actual values.

What is model validity?

The degree to which predictions match real-world outcomes.

What is model reliability?

The consistency of a model's performance across datasets.

What is the purpose of data visualization?

To clearly communicate the meaning of data through reports or presentations.

Why should visualization clutter be avoided?

It makes data harder to understand.

What should visualization titles include?

A neutral, factual description of what was measured and when.

What is the Gestalt principle of similarity?

Items similar in appearance are perceived as a group.

What is proximity in Gestalt theory?

Elements close together are perceived as related.

What is continuity in Gestalt theory?

Viewers follow lines or curves to interpret relationships.

What is closure in Gestalt theory?

The brain fills in incomplete shapes to form recognizable objects.

What is an omitted baseline?

Removing the zero baseline, making differences appear larger or smaller.

How does manipulating the y-axis mislead viewers?

It exaggerates or minimizes visual differences depending on scale.

What is an interactive visualization?

A visualization that allows users to explore data through filters or drill-downs.

What does a large language model do?

Understands and generates human-like text by learning from massive datasets.

How do LLMs help accountants?

Automating routine tasks, improving documentation, detecting risk, and interpreting standards.

What is prompt engineering?

Designing and refining prompts to obtain accurate responses from LLMs.

What is zero-shot prompting?

Asking a model to perform a task without examples.

What is few-shot prompting?

Providing a small number of examples in the prompt.

What is chain-of-thought prompting?

Encouraging step-by-step reasoning in the model's response.

What are hallucinations in LLMs?

Outputs that are factually incorrect, nonsensical, or unrelated to the input.