BISBA 513: Quantitative Analysis for Business - Day 1: Data Foundations

These flashcards cover data foundations, structuring, measurement scales, AI-assisted data management, and descriptive analysis based on the Day 1 lecture transcript of BISBA 513.

Data

A collection of facts obtained from experiences, observations, and measurements that are raw and unorganized.

Information

Data with context and meaning, occurring when data is organized, analyzed, and interpreted.

Knowledge

Information that provides a coherent, systematic understanding of a phenomenon.

Structured Data

Facts or observations with a high degree of organization that are searchable and come from a limited set of known categories or numerical scales.

Unstructured Data

Data that is usually not from a limited set of known values where every record may be unique and contain multiple facts, such as text, images, audio, and video.

Unit of Analysis (Observation)

The individual objects or subjects about which data is collected and analyzed, usually represented as a row in a dataset.

Variable

A measurable aspect of the phenomenon of interest, usually represented as the columns in a dataset.

Structured Output (AI)

Requiring an AI model to return fixed fields in a table format rather than prose.

Few-shot Prompting (AI)

Giving a model a handful of worked examples so it imitates human judgment instead of inventing its own standard.

Chain-of-thought (AI)

Asking an AI model to show its reasoning rather than just its answer to see why a case was coded a certain way.

Limited Context Window (AI)

The amount of text a model can hold or "remember" at once.

Model Drift (AI)

When the same prompt produces different results across batches, sessions, or model versions.

Hallucination (AI)

A fluent, confident output from an AI model that is factually made up.

Human-in-the-loop

A process where a person reviews and approves AI outputs at critical points.

Categorical Data

Data that represents groups or categories rather than quantitative information, consisting of nominal and ordinal scales.

Continuous Data

Numerical data that can take on a value between any other two values, consisting of ratio data and usually interval data.

Nominal Scale

A scale with two or more values that have no intrinsic order or numerical meaning.

Ordinal Scale

A scale where larger values indicate more (or less) of a characteristic, but not how much more (or less) in terms of specific magnitude.

Interval Scale

A scale where values are in a known order and are equidistant from the values before and after it.

Ratio Scale

A scale where distances between values are stated with respect to a rational or natural zero, representing $0$ quantity of the characteristic.

Discrete (Count) Data

A special category of numerical data that can only take on whole numbers, such as the number of children in a family.

Likert Scale

An interval scale (continuous) used as a proxy to measure constructs like satisfaction or unhappiness, typically using a range like $1$ to $5$.

Triangulation

Checking a conclusion against two or more measures to see if the story is consistent across measures that tap the same construct.

Behavioral Data

Observations of actions, such as clicks, eye fixations, or purchases.

Attitudinal Data

Data reflecting opinions or intentions, often measured on a Likert scale.

Descriptive Analysis

Analysis aimed at identifying trends and evaluating performance by describing the distribution of a variable (Who, What, When, and Where).

Correlational Analysis

Analysis used to determine if there is a relationship between variables, though it does not imply causation.

Causal Analysis

Analysis that establishes a cause-and-effect relationship, typically best tested with experiments.

JASP

A free, intuitive software used to run a wide range of statistical analyses, which can open $.csv$ files but not $.xlsx$ files.

Mean

The average value in a distribution; it can be misleading in skewed distributions or those with many zeros.

Median

The middle value in a distribution; it describes the typical case in skewed data (e.g., income, revenue, or wait times).

Bimodal Distribution

A distribution with two peaks where neither the mean nor the median is helpful, and groups should be reported separately.

Sensitivity Analysis (SA)

A "what-if" method used to examine what happens to results if the model, sample, measurements, or assumptions change, such as comparing results with and without outliers.

Bar Chart

A visualization used to present differences in magnitudes; vertical is preferred for key labels while horizontal is preferred for magnitude differences.

Histogram

A visualization used to show the distribution of continuous data and detect outliers.

Pie Chart

A visualization used only to show parts of a whole; the transcript advises avoiding them because they are often misleading.

Line Chart

A visualization used to show trends over time.

Scatter Plot

A visualization used to show correlation or identify outliers between variables.