EXIN Data Analytics Foundation flashcards

33 cards. Read the term and the definition here. The deck at the top of this page is optional.

Descriptive analytics

The analytics type that summarises historical data to answer what happened, using measures such as totals, averages, trends and dashboards. It forms the base for the other analytics types.

Prescriptive analytics

The analytics type that recommends actions to take, typically combining predictive output with optimisation or rules. It answers what should be done and usually requires human approval before execution.

Nominal scale

A categorical measurement scale whose values are labels with no order or ranking, such as country, product type or gender. Only frequency counts and the mode are meaningful statistics.

Ratio scale

A numerical scale with equal intervals and a true zero, such as revenue, height or age in years. Because zero means none, ratios between values are meaningful.

Discrete versus continuous data

Discrete data takes countable separate values, such as number of orders. Continuous data can take any value within a range, such as weight or duration, limited only by measurement precision.

Standard deviation

The square root of variance, expressing typical distance of values from the mean in the original unit. A low value means values cluster near the mean.

Correlation versus causation

Correlation means two variables move together linearly; causation means one produces change in the other. Correlation alone never proves causation because confounders or coincidence may explain it.

Structured data

Data organised in a predefined model of rows and columns, typically stored in relational databases and queried with SQL. It is easy to search, combine and aggregate.

Data analyst versus data scientist

A data analyst typically focuses on querying, cleaning and interpreting existing data for reporting and decisions. A data scientist additionally builds statistical and machine-learning models and often writes production code.

Problem definition step

The first step of the analysis process, converting a vague request into specific answerable questions, success criteria and constraints. A weak definition usually produces analysis nobody can act on.

Data cleaning

Preparing raw data for analysis by handling missing values and duplicates, correcting formats and types, standardising units and treating outliers. It usually consumes the largest share of analysis time.

Mean imputation

Replacing missing numerical values with the variable's mean. It preserves the sample mean but shrinks variance and can distort relationships, so it suits only small amounts of missing data.

IQR outlier rule

A value is flagged as an outlier when it falls below Q1 minus 1.5 times the interquartile range or above Q3 plus 1.5 times that range. The IQR is Q3 minus Q1.

Standardisation (z-score)

Rescaling a variable by subtracting its mean and dividing by its standard deviation, giving mean zero and standard deviation one. It lets variables with different units be compared or combined.

Stratified sampling

A sampling method that divides the population into homogeneous groups called strata, then draws random samples within each. It guarantees representation of every stratum and reduces sampling error.

Selection bias

A distortion caused when the sample is not representative of the population, for example because certain groups are more likely to be included or to respond. Conclusions then do not generalise.

Confirmation bias

The tendency to seek, interpret and remember evidence that supports existing beliefs while ignoring contradicting data. It is countered by pre-registering hypotheses and seeking disconfirming evidence.

Data validation

Checks that data meets defined rules during or after entry, such as type, range, format and referential integrity constraints. It prevents bad values entering the dataset rather than fixing them later.

A/B test

A controlled experiment randomly assigning subjects to a treatment and a control group, then comparing an outcome metric. Random assignment supports causal interpretation of the observed difference.

Bar chart use case

A chart for comparing values across discrete categories using bar length. Bars must start at zero, because a truncated baseline visually exaggerates differences between categories.

Scatter plot use case

A chart plotting two numerical variables as points to reveal relationship, direction, clusters and outliers. Adding a trend line shows the direction and strength of a linear association.

Histogram

A chart grouping one numerical variable into bins and showing the count or frequency per bin. It reveals distribution shape, central tendency, spread and possible skew.

Box plot

A chart showing median, quartiles, whiskers and flagged outliers for a numerical variable. It is efficient for comparing spread and skew across several groups side by side.

Sequential colour palette

A palette that varies in lightness or saturation of one hue to encode ordered low-to-high values, such as a heat map of sales. Diverging palettes suit data with a meaningful midpoint.

Data-ink ratio

Tufte's principle that the proportion of chart ink used to convey data should be maximised. It implies removing decoration, heavy gridlines and redundant labels that add no information.

Truncated axis

A chart axis that does not start at zero, exaggerating apparent differences. It is misleading for bar charts and area charts, whose size encodes magnitude, but acceptable for line charts.

Dashboard KPI placement

The convention of placing the most important indicator top-left, following reading order, and grouping related metrics with consistent units, time windows and a data-refresh timestamp.

Data governance

The framework of decision rights, policies and accountabilities that determines who may access and change data and who is responsible for its quality and security.

Data steward

The role responsible for day-to-day data quality, agreed definitions, metadata and resolving data issues within a domain. Stewards implement the standards that data owners are accountable for.

Data quality dimensions

Measurable properties of data fitness: completeness, accuracy, consistency, timeliness, uniqueness and validity. Each is tracked with metrics so quality can be monitored and improved.

Data minimisation

The GDPR principle that only personal data adequate, relevant and necessary for the stated purpose may be collected and processed. It limits exposure if a breach occurs.

Pseudonymisation

Replacing identifying fields with codes so data cannot be attributed to a person without separately held additional information. It reduces risk but the data remains personal data under GDPR.

Purpose limitation

The GDPR principle that personal data collected for a stated purpose may not be reused for incompatible purposes without a lawful basis. It constrains secondary analysis of existing datasets.