Conditional Probability & Screening - Business Statistics

1. A medical diagnostics company reports that its new screening test for a certain condition has a sensitivity of 95% and a Positive Predictive Value (PPV) of 80%. Which of the following pieces of information is also required to calculate the test's specificity?
   Answer: The prevalence of the condition in the population. Specificity is $P(\text{Test Negative} | \text{No Condition})$. The formula for PPV is $PPV = \frac{\text{Sens} \times \text{Prev}}{\text{Sens} \times \text{Prev} + (1-\text{Spec}) \times (1-\text{Prev})}$. We are given PPV (0.80) and Sensitivity (0.95). In the formula, both Prevalence (Prev) and Specificity (Spec) are unknown. To solve for Specificity, we must be given the Prevalence. The other options are not sufficient. Sample size or number of false negatives are outcomes, not parameters of the same type, and cost is irrelevant.
2. A financial fraud detection system flags 0.5% of legitimate transactions and fails to flag 5% of fraudulent transactions. If 0.1% of all transactions are fraudulent, what is the probability that a flagged transaction is actually fraudulent?
   Answer: Approximately 16.0% This is a classic Bayes' theorem problem in conditional probability. When you see questions about diagnostic tests, fraud detection, or any scenario involving false positives and false negatives, you need to work backwards from the test result to find the true probability. Let's set up the problem systematically. Define the events: F = fraudulent transaction, L = legitimate transaction, Flag = transaction is flagged. Given information: P(Flag|L) = 0.5% = 0.005, P(not Flag|F) = 5%, so P(Flag|F) = 95% = 0.95, and P(F) = 0.1% = 0.001, so P(L) = 99.9% = 0.999. Using Bayes' theorem: $P(F|Flag) = \frac{P(Flag|F) \times P(F)}{P(Flag)}$ First, find P(Flag) using the law of total probability: $P(Flag) = P(Flag|F) \times P(F) + P(Flag|L) \times P(L)$ $P(Flag) = 0.95 \times 0.001 + 0.005 \times 0.999 = 0.00095 + 0.004995 = 0.005945$ Now calculate: $P(F|Flag) = \frac{0.00095}{0.005945} ≈ 0.160 = 16.0\%$ Answer D is correct. Answer A (84%) likely comes from calculating P(Flag|F) instead of P(F|Flag). Answer B (95%) is the detection rate for fraudulent transactions, not the probability a flagged transaction is fraudulent. Answer C (0.1%) is simply the base rate of fraud, ignoring the flagging information entirely. Study tip: In Bayes' problems, the counterintuitive result often occurs when the base rate is very low—even accurate tests produce many false positives. Always distinguish between P(A|B) and P(B|A).
3. A performance review system categorizes employees as 'Below Expectations', 'Meets Expectations', or 'Exceeds Expectations'. In a given year, 15% of employees were rated 'Exceeds'. Of those rated 'Exceeds', 80% received a bonus. Of those not rated 'Exceeds', only 20% received a bonus. What is the probability that an employee who received a bonus was rated 'Exceeds Expectations'?
   Answer: 0.414 This is a Bayes' Theorem problem. Let E = 'Exceeds Expectations' and B = 'Received Bonus'. We want to find $P(E|B)$. We are given: $P(E) = 0.15$, so $P(E') = 0.85$. We are also given $P(B|E) = 0.80$ and $P(B|E') = 0.20$. First, we find the overall probability of receiving a bonus, $P(B)$, using the Law of Total Probability: $P(B) = P(B|E)P(E) + P(B|E')P(E') = (0.80)(0.15) + (0.20)(0.85) = 0.12 + 0.17 = 0.29$. Now, we can find $P(E|B) = \frac{P(B|E)P(E)}{P(B)} = \frac{(0.80)(0.15)}{0.29} = \frac{0.12}{0.29} \approx 0.4138$.
4. Following up on the previous scenario, suppose the company manufactures a different, low-value component where the primary concern is minimizing waste. The cost of a false positive (unnecessarily discarding a good component) is now considered more significant than the cost of a false negative (letting a flawed component pass). Using the same two tests (Test A: 99% sens, 90% spec; Test B: 95% sens, 98% spec), which test would be preferable in this new context?
   Answer: Test B, because its higher specificity minimizes the number of costly false positives. In this scenario, the cost of a false positive (discarding a good item) is the primary concern. Specificity is the probability that a good component is correctly identified as good ($P(\text{Test-}|\text{No Flaw})$). The false positive rate is $1 - \text{Specificity}$. To minimize false positives, the company needs the test with the highest specificity. Test B has a specificity of 98%, which is higher than Test A's 90%. Therefore, Test B is the better choice to minimize the waste of good components.
5. In a certain company, 40% of employees have a postgraduate degree. Of the employees with a postgraduate degree, 70% are in management positions. Of the employees without a postgraduate degree, 20% are in management positions. If an employee is selected at random and is found to be in a management position, what is the probability that they have a postgraduate degree?
   Answer: 0.70 Let PG = Postgraduate Degree and M = Management. We want to find $P(PG | M)$. We are given: $P(PG) = 0.40$, so $P(PG') = 0.60$. $P(M | PG) = 0.70$. $P(M | PG') = 0.20$. First, find the overall probability of being in management, $P(M)$, using the Law of Total Probability: $P(M) = P(M|PG)P(PG) + P(M|PG')P(PG') = (0.70)(0.40) + (0.20)(0.60) = 0.28 + 0.12 = 0.40$. Now use Bayes' Theorem: $P(PG|M) = \frac{P(M|PG)P(PG)}{P(M)} = \frac{(0.70)(0.40)}{0.40} = 0.70$.
6. A corporation sources a particular microchip from two suppliers: Alpha and Beta. 60% of the chips come from Alpha, and 40% come from Beta. The defect rate for chips from Alpha is 1%, while the defect rate for chips from Beta is 3%. If a microchip is selected at random from the inventory, what is the overall probability that it is defective?
   Answer: 0.018 This requires using the Law of Total Probability. Let D be the event a chip is defective, A be the event it's from Alpha, and B be the event it's from Beta. We want to find $P(D)$. We are given: $P(A) = 0.60$, $P(B) = 0.40$, $P(D|A) = 0.01$, and $P(D|B) = 0.03$. The formula is $P(D) = P(D|A)P(A) + P(D|B)P(B)$. Plugging in the values: $P(D) = (0.01)(0.60) + (0.03)(0.40) = 0.006 + 0.012 = 0.018$.
7. If quality control finds that a randomly selected pill is contaminated, what is the probability it was produced by Process B, and how does this compare to Process B's share of total production?
   Answer: 56.3%; this is 1.88 times higher than production share This question tests Bayes' theorem, which helps you find the probability of a cause given an observed effect. When you see quality control scenarios where you need to trace a defective item back to its source, think about how contamination rates and production volumes both matter. To find the probability that a contaminated pill came from Process B, you need to calculate what fraction of all contaminated pills originated from Process B. First, find the total contamination rate: $(0.50 × 0.01) + (0.30 × 0.03) + (0.20 × 0.02) = 0.005 + 0.009 + 0.004 = 0.018$ or 1.8%. Next, find what portion of contaminated pills came from Process B: $\frac{0.30 × 0.03}{0.018} = \frac{0.009}{0.018} = 0.563$ or 56.3%. Compared to Process B's 30% production share, this is $\frac{56.3\%}{30\%} = 1.88$ times higher. Choice A incorrectly calculates the probability as 18.8% (likely confusing it with the total contamination rate) and gets the wrong ratio. Choice B assumes the contamination probability equals the production share, ignoring that Process B has a higher contamination rate than average. Choice C gives just the contamination rate of Process B (3%), which isn't what the question asks for. Remember: when tracing defects back to their source, processes with higher defect rates will be overrepresented among the defective items compared to their production share. Always use Bayes' theorem to account for both the prior probability (production share) and the likelihood (contamination rate).
8. A screening test for a genetic condition has sensitivity 90% and specificity 95%. In a population where 2% carry the genetic condition, if 1000 people are tested, approximately how many of those who test negative actually have the condition?
   Answer: 2 people This question tests your understanding of conditional probability in diagnostic testing, specifically calculating false negatives. When analyzing screening tests, you need to work systematically through a 2×2 contingency table to avoid common calculation errors. Let's build this step by step with 1000 people tested. First, determine how many actually have the condition: $1000 \times 0.02 = 20$ people. Of these 20, the test correctly identifies $20 \times 0.90 = 18$ people (sensitivity = 90%). This means $20 - 18 = 2$ people with the condition test negative—these are false negatives. The question asks specifically for people who test negative but actually have the condition, which is exactly these 2 false negatives. Answer A (49 people) likely comes from incorrectly calculating total false positives instead of false negatives, or from misapplying the specificity rate. Answer B (18 people) represents the true positives—people correctly identified as having the condition. This is a common trap since 18 is prominently featured in our calculations, but it answers the wrong question. Answer D (931 people) represents the total number of people who test negative, including both false negatives and true negatives, but the question asks only for those who actually have the condition among the negative tests. When tackling diagnostic test problems, always create a complete 2×2 table showing true/false positives and negatives. This prevents confusion between similar-sounding quantities and ensures you're answering the specific question asked rather than calculating a related but incorrect value.
9. A medical screening test for a rare disease has a sensitivity of 92% and a specificity of 85%. The disease prevalence in the population is 0.3%. If a randomly selected individual tests positive, what is the probability they actually have the disease?
   Answer: Approximately 1.8% Using Bayes' theorem: P(Disease|Positive) = P(Positive|Disease) × P(Disease) / P(Positive). P(Positive) = 0.92 × 0.003 + 0.15 × 0.997 = 0.002760 + 0.14955 = 0.15231. Therefore P(Disease|Positive) = (0.92 × 0.003) / 0.15231 ≈ 0.018 or 1.8%. Choice B confuses sensitivity with posterior probability. Choice C confuses specificity with posterior probability. Choice D incorrectly uses P(Positive|Disease) / P(Positive|No Disease).
10. A manufacturer uses a screening test to detect defective products. The test has a sensitivity of 90% and a specificity of 95%. The true defect rate is 10%. If a product is tested and the result is negative (not defective), what is the probability that the product is actually not defective?
    Answer: 0.989 This question asks for the Negative Predictive Value (NPV), or $P(\text{Not Defective} | \text{Test Negative})$. Let D be the event a product is defective. We are given: $P(D) = 0.10$, Sensitivity $P(\text{Test +}|D) = 0.90$, Specificity $P(\text{Test -}|D') = 0.95$. From this, we can derive: $P(D') = 0.90$, False Negative Rate $P(\text{Test -}|D) = 1 - 0.90 = 0.10$. Using the formula for NPV: $NPV = \frac{P(\text{Test -}|D')P(D')}{P(\text{Test -}|D')P(D') + P(\text{Test -}|D)P(D)}$. Plugging in the values: $NPV = \frac{(0.95)(0.90)}{(0.95)(0.90) + (0.10)(0.10)} = \frac{0.855}{0.855 + 0.01} = \frac{0.855}{0.865} \approx 0.9884$.
11. A credit scoring model classifies loan applicants as 'low risk' or 'high risk'. A recent analysis showed that $P(\text{Default} | \text{Classified as High Risk}) = 0.40$. Which of the following is the correct interpretation of this statistic?
    Answer: Of the applicants that the model classifies as high risk, forty percent are expected to default. The notation $P(A|B)$ represents the conditional probability of event A occurring given that event B has already occurred. In this context, $P(\text{Default} | \text{Classified as High Risk}) = 0.40$ means that for the specific subgroup of applicants who were classified as high risk, there is a 40% probability of default. Choice C accurately describes this. Choice A describes $P(\text{Classified as High Risk} | \text{Default})$. Choice B describes the joint probability $P(\text{Classified as High Risk} \cap \text{Default})$. Choice D is an imprecise statement about accuracy; the statistic only concerns the default rate within the high-risk group, not the correctness of the classification itself for non-defaulters.
12. A loan approval algorithm has different error rates for different risk categories. For low-risk applicants (70% of all applicants), it incorrectly denies 3% of qualified candidates. For high-risk applicants (30% of all applicants), it incorrectly approves 12% of unqualified candidates. If 85% of low-risk and 40% of high-risk applicants are actually qualified, what is the probability that a randomly selected denied applicant was actually qualified?
    Answer: Approximately 40.4% Qualified applicants wrongly denied: Low-risk = 0.70 × 0.85 × 0.03 = 0.01785; High-risk = 0.30 × 0.40 × 0.88 = 0.1056. Unqualified applicants correctly denied: Low-risk = 0.70 × 0.15 × 1 = 0.105; High-risk = 0.30 × 0.60 × 1 = 0.18. Total denied = 0.01785 + 0.1056 + 0.105 + 0.18 = 0.40905. P(Qualified|Denied) = (0.01785 + 0.1056) / 0.40905 = 0.12345 / 0.40905 ≈ 0.404 or 40.4%. Choice A uses only the low-risk wrongly denied rate. Choice C gives P(Unqualified|Denied). Choice D uses the high-risk incorrect approval rate.
13. A financial services firm uses an algorithm to screen for fraudulent transactions. The algorithm has a sensitivity of 95% (it correctly identifies 95% of fraudulent transactions) and a specificity of 98% (it correctly identifies 98% of legitimate transactions). If the prevalence of fraudulent transactions in the system is 0.5%, what is the approximate probability that a transaction flagged as fraudulent by the algorithm is actually fraudulent?
    Answer: 0.192 This question asks for the Positive Predictive Value (PPV), which is $P(\text{Fraud} | \text{Flagged})$. Let F be the event of a fraudulent transaction and Flag be the event the algorithm flags it. We are given: $P(\text{Flag}|F) = 0.95$ (Sensitivity), $P(\text{Not Flag}|F') = 0.98$ (Specificity), and $P(F) = 0.005$ (Prevalence). From specificity, we find the false positive rate: $P(\text{Flag}|F') = 1 - 0.98 = 0.02$. We also know $P(F') = 1 - 0.005 = 0.995$. Using Bayes' Theorem: $P(F|\text{Flag}) = \frac{P(\text{Flag}|F)P(F)}{P(\text{Flag}|F)P(F) + P(\text{Flag}|F')P(F')}$. Plugging in the values: $P(F|\text{Flag}) = \frac{(0.95)(0.005)}{(0.95)(0.005) + (0.02)(0.995)} = \frac{0.00475}{0.00475 + 0.0199} = \frac{0.00475}{0.02465} \approx 0.1927$.
14. A cybersecurity firm develops a new system to detect malware. The system's sensitivity and specificity are held constant. If the system is deployed in a new environment where the prevalence of malware is significantly lower than the environment it was tested in, how will the Positive Predictive Value (PPV) of the system be affected?
    Answer: The PPV will decrease, because the ratio of false positives to true positives will increase. The Positive Predictive Value (PPV) is heavily dependent on the prevalence of the condition. The formula is $PPV = \frac{\text{Sens} \times \text{Prev}}{\text{Sens} \times \text{Prev} + (1-\text{Spec}) \times (1-\text{Prev})}$. As prevalence (Prev) decreases, the numerator ($\text{Sens} \times \text{Prev}$, representing true positives) decreases. The second term in the denominator ($(1-\text{Spec}) \times (1-\text{Prev})$, representing false positives) increases, as $(1-\text{Prev})$ gets larger. This means the proportion of positive tests that are true positives goes down, so the PPV decreases. A lower prevalence of malware means that a positive flag is more likely to be a false alarm.
15. An email spam filter has a 98% accuracy rate for spam emails and a 96% accuracy rate for legitimate emails. If 40% of all emails are spam, what is the probability that an email flagged as legitimate is actually spam?
    Answer: Approximately 9.1% P(Spam|Flagged Legitimate) = P(Flagged Legitimate|Spam) × P(Spam) / P(Flagged Legitimate). P(Flagged Legitimate|Spam) = 1 - 0.98 = 0.02. P(Flagged Legitimate) = 0.02 × 0.40 + 0.96 × 0.60 = 0.008 + 0.576 = 0.584. Therefore P(Spam|Flagged Legitimate) = (0.02 × 0.40) / 0.584 = 0.008 / 0.584 ≈ 0.091 or 9.1%. Choice A uses the false negative rate times spam prevalence. Choice C confuses the spam detection accuracy. Choice D uses the overall spam prevalence.
16. A company screens all new hires for a specific skill using a proficiency test. The test has a sensitivity of 92% and a specificity of 85%. Historically, 20% of applicants possess the skill. For a new hire who tests positive for the skill, what is the probability that they do not actually possess it?
    Answer: 0.395 This question asks for $P(\text{No Skill} | \text{Test Positive})$, which is the false positive rate among those who test positive, or $1 - \text{PPV}$. Let S = has skill. We have $P(S)=0.20$, $P(S')=0.80$, $P(\text{Test+}|S)=0.92$, and $P(\text{Test-}|S')=0.85$. The false positive rate is $P(\text{Test+}|S') = 1 - 0.85 = 0.15$. First, calculate PPV = $P(S|\text{Test+}) = \frac{P(\text{Test+}|S)P(S)}{P(\text{Test+}|S)P(S) + P(\text{Test+}|S')P(S')} = \frac{(0.92)(0.20)}{(0.92)(0.20) + (0.15)(0.80)} = \frac{0.184}{0.184 + 0.12} = \frac{0.184}{0.304} \approx 0.605$. The question asks for the probability they do not have the skill given a positive test, which is $1 - \text{PPV} = 1 - 0.605 = 0.395$.
17. An e-commerce company knows that 30% of its customers are 'high-value'. For these high-value customers, the probability of responding to a targeted email ad is 0.6. What is the probability that a randomly selected customer is both high-value AND responds to the ad?
    Answer: 0.18 This question asks for a joint probability, $P(A \cap B)$. Let A be the event that a customer is high-value, and B be the event they respond to the ad. We are given $P(A) = 0.30$ and the conditional probability $P(B|A) = 0.60$. The multiplication rule for conditional probability states $P(A \cap B) = P(B|A) \times P(A)$. Therefore, the probability is $0.60 \times 0.30 = 0.18$.
18. A company is considering two different tests to screen for a critical flaw in a high-value component. The cost of a false negative (failing to detect a flaw) is extremely high, leading to catastrophic failure. The cost of a false positive (discarding a good component) is significant but much lower. Test A has 99% sensitivity and 90% specificity. Test B has 95% sensitivity and 98% specificity. Which test should the company choose and why?
    Answer: Test A, because its higher sensitivity minimizes the very costly false negatives. The decision depends on the relative costs of the two types of errors. A false negative (a flawed component is passed as good) is described as having an 'extremely high' cost. Sensitivity is the probability that a truly flawed component is correctly identified ($P(\text{Test+}|\text{Flaw})$). A higher sensitivity means a lower false negative rate ($1 - \text{Sensitivity}$). Therefore, to minimize the most costly error, the company should choose the test with the highest sensitivity. Test A has 99% sensitivity, which is higher than Test B's 95%.
19. Using the information from the previous question (60% of chips from Alpha with a 1% defect rate, 40% from Beta with a 3% defect rate), suppose a chip is randomly selected and found to be defective. What is the probability that this defective chip came from supplier Beta?
    Answer: 0.667 This is a Bayes' Theorem problem. We want to find $P(\text{Beta} | \text{Defective})$. From the previous question, we calculated the overall probability of a defect, $P(D) = 0.018$. We need the term $P(D \cap \text{Beta})$, which is $P(D|\text{Beta})P(\text{Beta}) = (0.03)(0.40) = 0.012$. Now, apply the formula for conditional probability: $P(\text{Beta}|D) = \frac{P(D \cap \text{Beta})}{P(D)} = \frac{0.012}{0.018} = \frac{12}{18} = \frac{2}{3} \approx 0.667$.
20. In a quality control process, 3% of products are defective. A testing procedure correctly identifies 95% of defective products as defective, but also incorrectly classifies 8% of good products as defective. If a product is classified as defective by the test, what is the probability it is actually good?
    Answer: Approximately 73.1% P(Good|Test Defective) = P(Test Defective|Good) × P(Good) / P(Test Defective). P(Test Defective) = 0.95 × 0.03 + 0.08 × 0.97 = 0.0285 + 0.0776 = 0.1061. Therefore P(Good|Test Defective) = (0.08 × 0.97) / 0.1061 ≈ 0.731 or 73.1%. Choice B gives P(Defective|Test Defective). Choice C confuses the complement of the false positive rate. Choice D confuses the sensitivity of the test.

Start this quiz