DA0-002 — frequently asked questions

Masking hides or substitutes values for display and does not require a cryptographic key to reverse it, so it fails the question's "required key" criterion.

No. RBAC only controls who is permitted to access data; the data itself remains in readable form for anyone with the proper role.

Binning converts continuous values into ranges, such as score bands, for distribution analysis. It does not aggregate multiple rows per student, so it cannot produce a per-student average.

Each row represents a distinct score for that student, so repeated student names are legitimate records rather than duplicate data. Removing them would delete valid observations and skew the average.

A snapshot captures data at one fixed point in time, so every user sees the same stored data regardless of profile.

Real-time refers to how fresh the data is; dynamic refers to output that varies by user profile, filters, or parameters.

Data versioning stores successive snapshots for rollback and change history; it does not define standing guidelines or explain transformation rules in a readable reference.

A data dictionary defines data elements, standards, and transformation rules, while lineage traces where data originates and how it flows and changes between systems.

JSON keeps field names and types intact, so both sources produce DataFrames with matching columns. TXT and plain strings flatten the data, destroying the keys a merge needs.

varchar is a SQL column storage type, not an interchange format for Python. Casting API responses to varchar strips structure before pandas can build a usable DataFrame.

DISTINCT applies to the already-computed single total, not to the rows. The count still includes every job row, so duplicates remain.

Selecting JOB_TYPE splits the result into one row per job type instead of a single company-wide total, even though the syntax itself is valid.

Model bias means the model has a systematic, built-in error from flawed assumptions or unrepresentative training data. Here the model failed because demand changed after training, which is data drift, not an inherent modeling defect.

Skewing describes an uneven or distorted distribution inside a dataset. Data drift describes the mismatch between the training-time distribution and current real-world data — exactly what happens when a product suddenly goes viral.

Metadata provides a high-level summary (source, date), while a data dictionary details specific attributes (field names, types, definitions).

Option C describes metadata's function. A data dictionary goes beyond summary to define specific data elements and their structures.

Filters only restrict the dataset shown; they cannot calculate a metric at one granularity (region) while displaying it alongside another (salesperson).

No. Parameters accept user input values (like a number or text string) but do not perform mathematical aggregations on the underlying data.

JSON uses structural symbols like braces and brackets along with colons for keys, not a single character delimiter to separate all fields.

Yes, CSV can use semicolons, tabs, or pipes as delimiters, though commas are the most common standard.

Neither A nor B changes the dashboard itself, so visually impaired stakeholders would still be unable to read it. Accessibility must be designed into the deliverable, not referred elsewhere.

Option D replaces the requested dashboard with a different artifact, and ad-hoc sound effects do not provide structured screen-reader support. C applies audio accessibility consistently across the interface.

It is a one-time diagnostic action that may explain the current failure but provides no ongoing detection, so a future schema change would still break the dashboard unnoticed.

Alerts on data freshness, row/volume anomalies, null or duplicate spikes, and schema or field-type changes catch the common causes of dashboard breakage early.

Granularity refers to data detail levels during analysis, not the visual design of the survey instrument itself.

While important for web performance, it is not a primary accessibility consideration like color contrast when designing the survey content.

Outliers can drastically skew statistical measures like mean and standard deviation, leading to incorrect model training or insights.

The prompt mentions 'charts and tables,' which implies the structural type. The more significant gap is the logistical method of delivery (e.g., email vs. portal).

It belongs in data quality or acquisition phases. A report requirement focuses on the final output's appearance, content, and distribution.

A DFD shows how data moves through processes (flow), while an Entity-Relationship (ER) diagram shows how data is stored and related in a database structure (static relationships).

No, DFDs are for understanding system architecture and data movement. Auditing uses logs, traceability matrices, or specific compliance reports, not flow diagrams.

Redundancy is the state or condition of having duplicate copies available, which replication creates. The question asks for the practice itself, so replication is the more precise term.

Replication is a deliberate, controlled process of maintaining synchronized copies across systems or regions. Duplication usually means unintended or unnecessary extra copies that waste storage and can cause data-quality issues.

Organizing data for reporting describes structuring and preparing records before analysis. A pivot table's defining function is aggregating and analyzing that prepared data into summaries.

A pivot table outputs a summary table, not a chart. Dashboards consume pivot results through separate visualization objects, so visualization is not the pivot table's purpose.

Variance does measure dispersion, but it is expressed in squared units of the data, making it hard to interpret. Standard deviation, its square root, returns to the original units and is the measure CompTIA expects for describing distribution.

A t-test is an inferential test that compares the means of two groups to check for a statistically significant difference. It does not describe how scores within a single class are spread around their mean.

Joins and merges combine tables horizontally on matching keys, adding columns side by side. To add missing months you need more rows, which is what appending does.

Parse means interpreting or splitting structured text, such as separating a date string into parts. It does not retrieve missing records from a data source.

It is costly, may be blocked by environment limits, and only masks the problem. Chunking fixes the ingestion method itself and works with existing resources.

It applies broadly — CSV files, log streams, and ETL pipelines are all processed in smaller batches or with streaming readers to avoid exhausting memory.

Missing values require actual blank or null cells in required fields; here every field is populated and the defect is an identical repeated row.

Duplication repeats whole records inside one table, while redundancy stores the same attribute in multiple columns or tables — only repeated records appear in this exhibit.

The mode is 37 because it occurs more frequently than the other listed values; the mode is defined by the highest frequency, not by the average or middle position.

A repeated value is only the mode if no other value occurs more often. In this exhibit, 37 has the higher frequency, so 41 is a distractor based on incomplete counting.

Yes. If two or more values tie for the highest frequency, each is a mode; if all values occur equally often, the data set has no mode. This question's exhibit has a single highest-frequency value.