Back to CompTIA Data+ V2 (Data+)

Data Analysis

Apply statistical methods and analytical techniques to extract insights from data and communicate findings effectively.

Covers communicating analysis results by selecting appropriate methods for different audiences and stakeholders. Includes selecting and applying statistical methods such as descriptive statistics, inferential statistics, and basic statistical techniques to analyze data. Also covers troubleshooting analysis issues by using appropriate tools and resources to identify and resolve problems in data analysis workflows.
5 minutes 5 Questions

In the context of the CompTIA Data+ certification (Data+ V2), Data Analysis is defined as the systematic process of inspecting, cleansing, transforming, and modeling data with the goal of discovering useful information, informing conclusions, and supporting decision-making. Unlike advanced data sci…

Concepts covered: Tailoring communication to audiences, Technical vs. non-technical reporting, Storytelling with data, Data-driven recommendations, Presentation best practices, Descriptive statistics, Measures of central tendency (mean, median, mode), Measures of dispersion (variance, standard deviation), Correlation and regression basics, Hypothesis testing fundamentals, Statistical significance and p-values, Sampling methods and techniques, Trend analysis and forecasting, A/B testing and experimentation, Debugging data analysis workflows, Identifying data quality issues, Resolving calculation errors, Using documentation and resources, Version control for data analysis, Collaborative troubleshooting, Executive summaries and key findings

Test mode:
Start practice test
Data+ - Data Analysis Example Questions

Test your knowledge of Data Analysis

Question 1

A data analyst is evaluating customer transaction amounts from 400 records. The calculated z-scores reveal that 2.8% of transactions fall beyond ±2.2 standard deviations from the mean, while the empirical rule predicts approximately 2.8% should fall beyond ±2.2 standard deviations for a normal distribution. The skewness is 0.12 and kurtosis is 3.08. Additionally, when the analyst creates a probability plot, the points align closely with the diagonal reference line. However, a Shapiro-Wilk test yields a p-value of 0.04. Given these conflicting indicators with a significance level of 0.05, which conclusion demonstrates the most sophisticated understanding of normality assessment?

Correct Answer: The formal hypothesis test rejects normality at alpha 0.05, but practical normality is supported by the descriptive statistics and visual assessment, suggesting the data is approximately normal for most analytical purposes

This question tests understanding of normality assessment in data analysis, which requires balancing multiple diagnostic tools rather than relying on a single test.

The correct answer demonstrates sophisticated statistical reasoning by recognizing that normality assessment should consider multiple indicators holistically:

  1. The z-score distribution (2.8% beyond ±2.2 SD) matches theoretical normal distribution expectations
  2. Skewness of 0.12 is very close to 0 (indicating near-symmetric distribution)
  3. Kurtosis of 3.08 is nearly identical to the theoretical normal value of 3
  4. The probability plot shows close alignment with the reference line
  5. The Shapiro-Wilk test, while statistically significant at p=0.04, represents just one piece of evidence

With 400 records, formal hypothesis tests like Shapiro-Wilk become very sensitive and can detect trivial deviations from perfect normality that have no practical significance. This is the concept of statistical significance versus practical significance. When all descriptive and visual measures suggest normality, but only the hypothesis test (barely) rejects it, the sophisticated approach recognizes that 'approximate normality' is sufficient for most analytical purposes.

The second option is incorrect because it takes an overly rigid approach, treating the Shapiro-Wilk test as definitive when multiple other indicators support normality. Statistical significance at p=0.04 with n=400 may reflect minor, inconsequential deviations.

The third option incorrectly claims the values 'equal' theoretical values - they are close but not equal (0.12 ≠ 0, 3.08 ≠ 3). It also incorrectly assumes the Shapiro-Wilk result must be a Type I error, which cannot be determined.

The fourth option makes two errors: no single test overrides all others in normality assessment, and a skewness of 0.12 (essentially zero) does not suggest lognormal distribution. Lognormal data typically shows substantial positive skewness, not near-zero skewness.

Question 2

A customer insights team and a web analytics group have been collaboratively troubleshooting why user conversion funnels show different completion rates for six days. The customer insights team uses event-based tracking from the mobile app SDK, while the web analytics group relies on server-side session logs. Both teams have validated their individual tracking implementations and confirmed data collection is functioning correctly. During joint meetings, each team presents summaries of their findings but tends to defend their own methodology as the accurate baseline. A breakthrough is needed to move past this impasse. Which collaborative troubleshooting strategy would most effectively help both teams identify where their data interpretations fundamentally diverge?

Correct Answer: Create a shared test scenario with controlled user actions and trace the same events through both tracking systems simultaneously to compare outputs at each stage

The correct answer is to create a shared test scenario with controlled user actions and trace the same events through both tracking systems simultaneously to compare outputs at each stage.

This approach is the most effective collaborative troubleshooting strategy because:

  1. Common Ground Testing: By using identical, controlled user actions, both teams work from the exact same input data, eliminating variables and creating a true apples-to-apples comparison.

  2. Pinpoints Divergence: Tracing the same events through both systems simultaneously allows teams to identify precisely WHERE the data interpretations begin to differ - whether it's at data collection, processing, aggregation, or reporting stages.

  3. Objectivity: This removes the defensive posture of each team because the focus shifts from defending methodologies to observing objective results from controlled conditions.

  4. Actionable Insights: Finding the specific stage where outputs diverge provides concrete information to resolve the discrepancy.

Why the other options are less effective:

  • Exchanging documentation and reviewing methodology summaries continues the same pattern that has already failed - both teams have been presenting summaries in meetings. More documentation review doesn't create new insights about where data actually diverges.

  • A rotating facilitation schedule where teams explain why their approach is more reliable perpetuates the adversarial dynamic and defensive posturing. It doesn't address the root cause of divergence.

  • Building a comparison matrix to determine which system should be authoritative misses the point - the goal is to understand WHY the systems show different results, not to pick a winner. Understanding the differences is necessary before determining appropriate use cases for each system.

Question 3

A data analyst is debugging a real-time dashboard that displays customer churn predictions. The dashboard pulls data from a stored procedure that joins five tables and applies a machine learning model's probability scores. Users report that churn probabilities occasionally display as negative values, which should be mathematically impossible given the model's sigmoid output function. The stored procedure executes with no errors, the ML model service returns valid responses, and database constraints show no violations. Network latency metrics are within acceptable ranges. Which debugging strategy would most effectively identify the source of these impossible negative probability values?

Correct Answer: Trace the data flow by logging intermediate values at each transformation step within the stored procedure to identify where valid probabilities become corrupted

The most effective debugging strategy is to trace the data flow by logging intermediate values at each transformation step within the stored procedure. This approach is the gold standard for debugging data corruption issues because it allows you to pinpoint exactly where valid data becomes invalid.

Since the ML model's sigmoid function guarantees outputs between 0 and 1, and the model service is confirmed to return valid responses, the corruption must be happening somewhere in the data pipeline after the ML service returns its values. By instrumenting each transformation step with logging, you can observe the probability values as they move through the stored procedure's five table joins and any calculations. This will reveal the exact point where positive probabilities transform into negative values.

The second option about data type conversions and numeric overflow is a reasonable consideration, but it's a more targeted hypothesis rather than a systematic debugging approach. While overflow could potentially cause sign changes, this strategy assumes the problem before gathering evidence.

The third option about join conditions causing calculation errors during aggregation is less likely because join mismatches typically result in NULL values or missing records rather than mathematically impossible negative values from a sigmoid function.

The fourth option about caching mechanisms retrieving stale values doesn't explain the appearance of negative probabilities. Stale cached values would still be valid probabilities from previous calculations - they would be outdated but not negative, since the ML model always produces values between 0 and 1.

The systematic logging approach provides comprehensive visibility into the entire data transformation pipeline, making it the most effective method to identify where and why the corruption occurs.

More Data Analysis questions
25 questions (total)
Start 25 question test