CompTIA DY0-001 Practice Test

DY0-001 practice test: what's covered

• Statistics
• Modeling
• Machine Learning
• MLOps
• Deep Learning

Free DY0-001 sample questions

1. A data scientist models the number of fraudulent transactions arriving per hour at a payment gateway, where events are rare and independent and the mean rate is roughly constant. Which probability distribution best represents this count process?

   • A. Poisson distribution
   • B. Binomial distribution
   • C. Exponential distribution
   • D. Beta distribution
   Show answer

   Correct: A. The Poisson distribution models the count of rare, independent events over a fixed interval given a constant mean rate, which exactly fits hourly fraud counts. The binomial is for a fixed number of trials with success probability, not an open-ended rate. The exponential models waiting time between events, not the count itself, and the beta models probabilities on [0,1].

2. An analyst runs an A/B test and obtains a p-value of 0.03 at a significance level of 0.05. A stakeholder claims this means there is a 3% chance the null hypothesis is true. How should the data scientist correct this interpretation?

   • A. The p-value is the probability of data this extreme assuming the null is true
   • B. The p-value equals the probability that the null hypothesis is correct
   • C. The p-value is the probability the alternative hypothesis is false
   • D. The p-value measures the effect size of the observed difference
   Show answer

   Correct: A. A p-value is the probability of observing a result at least as extreme as the data, conditioned on the null being true, not the probability that the null is true. The stakeholder's reading inverts the conditioning, a classic fallacy. The p-value says nothing directly about the alternative being false, and it is not an effect-size measure because it depends heavily on sample size.

3. A data scientist must choose appropriate corrections and methods when running 50 simultaneous hypothesis tests on gene-expression features to control false discoveries. Which approaches validly address the multiple-comparisons problem? (Choose TWO)

   • A. Apply the Bonferroni correction to the significance threshold
   • B. Use the Benjamini-Hochberg procedure to control the false discovery rate
   • C. Increase the per-test alpha to 0.10 to gain power
   • D. Report only the single smallest p-value without adjustment
   • E. Switch from two-tailed to one-tailed tests to lower p-values
   Show answer

   Correct: A and B. Bonferroni controls the family-wise error rate by dividing alpha across tests, and Benjamini-Hochberg controls the expected false discovery rate, both valid multiple-comparison corrections. Raising alpha increases false positives rather than controlling them. Cherry-picking the smallest p-value ignores the multiplicity entirely, and switching tail direction does not address the inflated family-wise error from running 50 tests.

4. A medical screening test has 99% sensitivity and 95% specificity, but the disease prevalence in the screened population is only 0.5%. A patient tests positive and asks for the probability they actually have the disease. Which concept must the data scientist apply to answer correctly?

   • A. Bayes' theorem combining prior prevalence with the likelihood
   • B. The central limit theorem applied to the test scores
   • C. The law of large numbers over repeated screenings
   • D. A simple ratio of sensitivity to specificity
   Show answer

   Correct: A. Bayes' theorem updates the low prior (0.5% prevalence) with the test likelihoods, and with such a rare condition the posterior probability of disease given a positive remains low despite high sensitivity. The central limit theorem and law of large numbers describe sampling distributions, not posterior updating. A raw ratio of sensitivity to specificity ignores the base rate, which is precisely what drives the counterintuitive result.

5. While preparing features for a linear model, a data scientist finds a predictor whose histogram is strongly right-skewed with a long tail of large positive values, violating normality assumptions for residuals. Which transformation is the most appropriate first remedy?

   • A. Apply a log transformation to compress the right tail
   • B. Standardize the feature to zero mean and unit variance
   • C. One-hot encode the continuous values into bins
   • D. Replace all values with their rank percentile only
   Show answer

   Correct: A. A log transformation compresses large positive values and pulls in a right-skewed tail, making the distribution more symmetric and stabilizing variance. Standardization only shifts and scales but preserves skew, so it does not fix the shape. One-hot encoding is for categorical data and discards continuous information, and rank-only replacement throws away magnitude that the linear model needs.

Last updated: 2026-10-19

DY0-001 practice test FAQ

Related practice tests

• CompTIA A+ 220-1101 / 220-1102 practice test
• CompTIA CAS-005 practice test
• CompTIA CLO-002 practice test
• CompTIA CV0-004 practice test
• CySA+ CS0-003 practice test
• CompTIA A+ DA0-001 practice test