Binomial vs Geometric Distribution: How to Choose
Choose the correct count or waiting-time model by reading the stopping rule first, while keeping geometric distribution clearly labeled as legacy content in the revised AP Statistics course.
Binomial Vs Geometric Distribution: direct answer
Use this later comparison checkpoint to read the stopping rule before touching the calculator: a fixed number of trials signals binomial reasoning, whereas waiting until a first success describes the legacy geometric model. For 2027 preparation, treat the geometric side only as enrichment and keep current exam practice centered on the binomial model.
This page uses binomial vs geometric distribution as its single primary search focus. Every instructional block and every retained practice item is tied to that title intent rather than to a generic AP Statistics question-bank template.
Quick reference for Binomial vs Geometric Distribution: How to Choose
| Fixed n | Binomial counts successes in a predetermined number of trials. |
|---|---|
| Stop at first success | Geometric waiting time is a legacy comparison, removed from the revised tested course. |
| Binomial mean | μ=np |
| Binomial SD | σ=√[np(1−p)] |
| Legacy geometric mean | 1/p |
Concept mastery: Binomial vs Geometric Distribution: How to Choose
The stopping rule is the fastest classifier
A fixed trial count points toward a binomial count; stopping at the first success points toward the historical geometric waiting-time model. The two models may share independent Bernoulli trials and a constant p, but they ask different random-variable questions.
Why geometric is legacy rather than current core
College Board removed geometric distribution from the revised AP Statistics course effective in 2026–27. A comparison remains useful when students encounter older textbooks or archived practice, but current exam preparation should not allocate geometric distribution the same priority as binomial probability.
Binomial calculations use combinations
For exactly k successes among n trials, the binomial formula multiplies the probability of one success/failure ordering by C(n,k), the number of orderings containing k successes. A first-success waiting-time calculation has no combination factor because the last trial is fixed as the first success and every preceding trial is a failure.
Expected values answer different questions
For X~Bin(n,p), E(X)=np counts average successes per n-trial batch. For the legacy first-success model, E(X)=1/p counts average trials until success. Equal numerical means would not make the random variables interchangeable because their supports and stopping rules differ.
Independence and constant probability still matter
Both simple models assume a stable success probability and independent or approximately independent trials. Learning, depletion, fatigue, replacement rules, or adaptive difficulty can change p and invalidate both models even when every trial can be labeled success or failure.
Language cues are evidence, not proof
Phrases such as “out of 20” or “among 50” often signal a fixed n, while “until the first” signals a waiting time. Always translate the random variable explicitly because surface wording can be misleading in more complicated designs.
Model-choice atlas: fixed campaign versus first-success search
A fixed campaign finishes all planned trials and then counts successes, even if a success occurs immediately. A first-success search stops as soon as the target event appears. That operational difference determines the random variable before any probability is calculated. For example, sending 40 emails and counting replies is a fixed-n count; sending emails one by one until the first reply is a waiting-time process. The same underlying reply probability can therefore lead to different models because the stopping rule and recorded response are different.
Model-choice atlas: why a maximum trial limit can be misleading
A process may allow at most 20 attempts but stop after the first success. The existence of a maximum does not turn the observed waiting time into a binomial count. If the recorded variable is the attempt on which first success occurs, the historical waiting-time logic is still the relevant comparison, possibly with truncation at 20. Conversely, if all 20 attempts are completed and the recorded variable is the number of successes, the count is binomial when the other conditions hold.
Model-choice atlas: finite-population sampling
Without-replacement sampling deserves a separate check because each selection changes the remaining population. When the sample is a small fraction of a large population, the change can be negligible enough for an approximate binomial model. When the sample consumes a large fraction, independence and constant-p assumptions deteriorate. This is a design issue, not a calculation issue: no combination formula can restore assumptions that the sampling mechanism violates.
Model-choice atlas: adaptive systems
Adaptive systems deliberately alter the next trial after observing the current result. A learning platform may increase difficulty after success, a machine may heat up after repeated operation, or a salesperson may change the pitch after rejection. In each case, p can change across trials. Binary outcomes remain present, but the simple constant-p binomial and historical geometric models are no longer automatically justified. The correct response is to describe the changing mechanism rather than force a familiar distribution.
Model-choice atlas: grouped or clustered observations
Trials can be binary yet dependent because they share a cluster. Household members, products from the same machine cycle, or patients treated by the same clinic may respond similarly. A fixed n does not erase that correlation. Before using a binomial model, ask whether the repeated Bernoulli trials are approximately independent at the level being analyzed. If clustering is substantial, the usual binomial variance can understate the true sampling variability.
Model-choice atlas: event boundaries after the model is chosen
Model choice and event translation are separate decisions. In a binomial count, “at least 5” means X≥5 and is usually computed as 1−P(X≤4); “more than 5” means X≥6. In a first-success waiting-time problem, “more than 5 trials” means the first five trials are failures. The same English phrase therefore maps to a different probability structure because the random variables have different meanings.
Model-choice atlas: expectation without determinism
The binomial expectation np is the long-run average number of successes per fixed-size batch. The historical geometric expectation 1/p is the long-run average trial number of first success. Neither value is a guarantee for one repetition. A batch with expected count 4 can produce 0, 1, 7, or another feasible count, while a waiting-time process with expected value 5 can succeed on trial 1 or continue far beyond trial 5.
Model-choice atlas: current-course study priority
For the revised AP Statistics course, binomial distribution remains current content while geometric distribution was removed. The comparison is still valuable because it teaches students to define a random variable and inspect a stopping rule, but the study plan should reflect the current framework. Use the legacy model as a contrast tool when reading older materials, then return most practice time to binomial probability, sampling distributions, data collection, inference, and regression.
Worked analysis for Binomial vs Geometric Distribution: How to Choose
Decision case 1: Quality Audit
Design. The process is to inspect 18 devices and count failures with success probability 0.08 on each independent trial.
Model decision. This is binomial because the trial count is fixed before inspection begins. Before touching a formula, identify whether the random variable counts successes inside a fixed n or counts trials until the first success.
Interpretation. If the independence and constant-p assumptions are defensible, the long-run count has mean 1.44 and standard deviation 1.15. Those summaries describe repeated 18-trial batches; they do not describe the probability of a single trial.
Decision case 2: Fundraising Calls
Design. The process is to call donors until the first pledge arrives with success probability 0.24 on each independent trial.
Model decision. This is geometric because the stopping time is the random variable. Before touching a formula, identify whether the random variable counts successes inside a fixed n or counts trials until the first success.
Interpretation. Under the historical geometric model, the expected trial number of the first success is 1/p≈4.17. That waiting-time interpretation is fundamentally different from a binomial count, even though both models use two outcome categories and a constant p.
Decision case 3: Free-Throw Drill
Design. The process is to take exactly 12 shots and count makes with success probability 0.71 on each independent trial.
Model decision. This is binomial because successes are counted inside a fixed set of trials. Before touching a formula, identify whether the random variable counts successes inside a fixed n or counts trials until the first success.
Interpretation. If the independence and constant-p assumptions are defensible, the long-run count has mean 8.52 and standard deviation 1.57. Those summaries describe repeated 12-trial batches; they do not describe the probability of a single trial.
Decision case 4: Server Restart
Design. The process is to repeat a restart protocol until the first successful boot with success probability 0.36 on each independent trial.
Model decision. This is geometric because trials continue until first success. Before touching a formula, identify whether the random variable counts successes inside a fixed n or counts trials until the first success.
Interpretation. Under the historical geometric model, the expected trial number of the first success is 1/p≈2.78. That waiting-time interpretation is fundamentally different from a binomial count, even though both models use two outcome categories and a constant p.
Decision case 5: Seed Germination
Design. The process is to plant 30 seeds and count how many germinate with success probability 0.83 on each independent trial.
Model decision. This is binomial because n=30 is fixed at the design stage. Before touching a formula, identify whether the random variable counts successes inside a fixed n or counts trials until the first success.
Interpretation. If the independence and constant-p assumptions are defensible, the long-run count has mean 24.90 and standard deviation 2.06. Those summaries describe repeated 30-trial batches; they do not describe the probability of a single trial.
Decision case 6: Sales Outreach
Design. The process is to contact prospects until the first booked demonstration with success probability 0.12 on each independent trial.
Model decision. This is geometric because the number of contacts is not fixed. Before touching a formula, identify whether the random variable counts successes inside a fixed n or counts trials until the first success.
Interpretation. Under the historical geometric model, the expected trial number of the first success is 1/p≈8.33. That waiting-time interpretation is fundamentally different from a binomial count, even though both models use two outcome categories and a constant p.
Decision case 7: Screening Batch
Design. The process is to screen 25 records and count records with an error with success probability 0.06 on each independent trial.
Model decision. This is binomial because the count is successes out of 25. Before touching a formula, identify whether the random variable counts successes inside a fixed n or counts trials until the first success.
Interpretation. If the independence and constant-p assumptions are defensible, the long-run count has mean 1.50 and standard deviation 1.19. Those summaries describe repeated 25-trial batches; they do not describe the probability of a single trial.
Decision case 8: Machine Calibration
Design. The process is to repeat calibration attempts until the first pass with success probability 0.42 on each independent trial.
Model decision. This is geometric because X records trial number of first pass. Before touching a formula, identify whether the random variable counts successes inside a fixed n or counts trials until the first success.
Interpretation. Under the historical geometric model, the expected trial number of the first success is 1/p≈2.38. That waiting-time interpretation is fundamentally different from a binomial count, even though both models use two outcome categories and a constant p.
Decision case 9: Survey Response
Design. The process is to send 40 invitations and count completed surveys with success probability 0.58 on each independent trial.
Model decision. This is binomial because there are exactly 40 Bernoulli trials. Before touching a formula, identify whether the random variable counts successes inside a fixed n or counts trials until the first success.
Interpretation. If the independence and constant-p assumptions are defensible, the long-run count has mean 23.20 and standard deviation 3.12. Those summaries describe repeated 40-trial batches; they do not describe the probability of a single trial.
Decision case 10: Penalty Kicks
Design. The process is to take attempts until the first goal with success probability 0.67 on each independent trial.
Model decision. This is geometric because the first success ends the process. Before touching a formula, identify whether the random variable counts successes inside a fixed n or counts trials until the first success.
Interpretation. Under the historical geometric model, the expected trial number of the first success is 1/p≈1.49. That waiting-time interpretation is fundamentally different from a binomial count, even though both models use two outcome categories and a constant p.
Decision case 11: Delivery Scan
Design. The process is to check 20 packages and count late arrivals with success probability 0.11 on each independent trial.
Model decision. This is binomial because fixed n and success count define the variable. Before touching a formula, identify whether the random variable counts successes inside a fixed n or counts trials until the first success.
Interpretation. If the independence and constant-p assumptions are defensible, the long-run count has mean 2.20 and standard deviation 1.40. Those summaries describe repeated 20-trial batches; they do not describe the probability of a single trial.
Decision case 12: Diagnostic Retest
Design. The process is to repeat independent tests until the first positive with success probability 0.18 on each independent trial.
Model decision. This is geometric because waiting time to first success is modeled. Before touching a formula, identify whether the random variable counts successes inside a fixed n or counts trials until the first success.
Interpretation. Under the historical geometric model, the expected trial number of the first success is 1/p≈5.56. That waiting-time interpretation is fundamentally different from a binomial count, even though both models use two outcome categories and a constant p.
Decision case 13: Email Campaign
Design. The process is to send exactly 50 messages and count replies with success probability 0.09 on each independent trial.
Model decision. This is binomial because the campaign fixes n=50. Before touching a formula, identify whether the random variable counts successes inside a fixed n or counts trials until the first success.
Interpretation. If the independence and constant-p assumptions are defensible, the long-run count has mean 4.50 and standard deviation 2.02. Those summaries describe repeated 50-trial batches; they do not describe the probability of a single trial.
Decision case 14: Game Loot
Design. The process is to open boxes until the first rare item appears with success probability 0.03 on each independent trial.
Model decision. This is geometric because the number opened before first rare item is random. Before touching a formula, identify whether the random variable counts successes inside a fixed n or counts trials until the first success.
Interpretation. Under the historical geometric model, the expected trial number of the first success is 1/p≈33.33. That waiting-time interpretation is fundamentally different from a binomial count, even though both models use two outcome categories and a constant p.
Decision case 15: Manufacturing Lot
Design. The process is to sample 15 units and count cosmetic defects with success probability 0.14 on each independent trial.
Model decision. This is binomial because the statistic is the number of successes in a fixed sample. Before touching a formula, identify whether the random variable counts successes inside a fixed n or counts trials until the first success.
Interpretation. If the independence and constant-p assumptions are defensible, the long-run count has mean 2.10 and standard deviation 1.34. Those summaries describe repeated 15-trial batches; they do not describe the probability of a single trial.
Decision case 16: Password Reset
Design. The process is to retry authentication until the first accepted attempt with success probability 0.55 on each independent trial.
Model decision. This is geometric because the random variable is the attempt of first success. Before touching a formula, identify whether the random variable counts successes inside a fixed n or counts trials until the first success.
Interpretation. Under the historical geometric model, the expected trial number of the first success is 1/p≈1.82. That waiting-time interpretation is fundamentally different from a binomial count, even though both models use two outcome categories and a constant p.
Decision case 17: Vaccination Reminder
Design. The process is to text 32 patients and count confirmations with success probability 0.63 on each independent trial.
Model decision. This is binomial because the number contacted is fixed. Before touching a formula, identify whether the random variable counts successes inside a fixed n or counts trials until the first success.
Interpretation. If the independence and constant-p assumptions are defensible, the long-run count has mean 20.16 and standard deviation 2.73. Those summaries describe repeated 32-trial batches; they do not describe the probability of a single trial.
Decision case 18: Field Sensor
Design. The process is to replace batteries until the first sensor passes diagnostics with success probability 0.47 on each independent trial.
Model decision. This is geometric because sampling stops at the first pass. Before touching a formula, identify whether the random variable counts successes inside a fixed n or counts trials until the first success.
Interpretation. Under the historical geometric model, the expected trial number of the first success is 1/p≈2.13. That waiting-time interpretation is fundamentally different from a binomial count, even though both models use two outcome categories and a constant p.
Decision case 19: Quiz Guessing
Design. The process is to answer 10 independent four-choice items and count correct answers with success probability 0.25 on each independent trial.
Model decision. This is binomial because ten trials are completed regardless of earlier outcomes. Before touching a formula, identify whether the random variable counts successes inside a fixed n or counts trials until the first success.
Interpretation. If the independence and constant-p assumptions are defensible, the long-run count has mean 2.50 and standard deviation 1.37. Those summaries describe repeated 10-trial batches; they do not describe the probability of a single trial.
Decision case 20: Support Escalation
Design. The process is to route cases until the first case requiring escalation with success probability 0.16 on each independent trial.
Model decision. This is geometric because waiting time rather than a fixed-trial count is observed. Before touching a formula, identify whether the random variable counts successes inside a fixed n or counts trials until the first success.
Interpretation. Under the historical geometric model, the expected trial number of the first success is 1/p≈6.25. That waiting-time interpretation is fundamentally different from a binomial count, even though both models use two outcome categories and a constant p.
Decision case 21: Mail Inspection
Design. The process is to inspect exactly 28 envelopes and count damaged seals with success probability 0.07 on each independent trial.
Model decision. This is binomial because all 28 trials occur whether or not early damage is found. Before touching a formula, identify whether the random variable counts successes inside a fixed n or counts trials until the first success.
Interpretation. If the independence and constant-p assumptions are defensible, the long-run count has mean 1.96 and standard deviation 1.35. Those summaries describe repeated 28-trial batches; they do not describe the probability of a single trial.
Decision case 22: Safety Alarm
Design. The process is to test sensors sequentially until the first alarm activates with success probability 0.31 on each independent trial.
Model decision. This is geometric because testing stops at the first success. Before touching a formula, identify whether the random variable counts successes inside a fixed n or counts trials until the first success.
Interpretation. Under the historical geometric model, the expected trial number of the first success is 1/p≈3.23. That waiting-time interpretation is fundamentally different from a binomial count, even though both models use two outcome categories and a constant p.
Decision case 23: Scholarship Outreach
Design. The process is to email 45 applicants and count accepted interviews with success probability 0.22 on each independent trial.
Model decision. This is binomial because the number of outreach attempts is fixed at 45. Before touching a formula, identify whether the random variable counts successes inside a fixed n or counts trials until the first success.
Interpretation. If the independence and constant-p assumptions are defensible, the long-run count has mean 9.90 and standard deviation 2.78. Those summaries describe repeated 45-trial batches; they do not describe the probability of a single trial.
Decision case 24: Robot Restart
Design. The process is to restart the controller until the first error-free cycle with success probability 0.41 on each independent trial.
Model decision. This is geometric because the first successful cycle determines the stopping time. Before touching a formula, identify whether the random variable counts successes inside a fixed n or counts trials until the first success.
Interpretation. Under the historical geometric model, the expected trial number of the first success is 1/p≈2.44. That waiting-time interpretation is fundamentally different from a binomial count, even though both models use two outcome categories and a constant p.
Decision case 25: Voter Contact
Design. The process is to dial exactly 60 selected numbers and count completed interviews with success probability 0.37 on each independent trial.
Model decision. This is binomial because the count of completed interviews comes from a fixed contact list. Before touching a formula, identify whether the random variable counts successes inside a fixed n or counts trials until the first success.
Interpretation. If the independence and constant-p assumptions are defensible, the long-run count has mean 22.20 and standard deviation 3.74. Those summaries describe repeated 60-trial batches; they do not describe the probability of a single trial.
Decision case 26: Medical Screening
Design. The process is to repeat screening on independent samples until the first reactive result with success probability 0.09 on each independent trial.
Model decision. This is geometric because the trial number of first reaction is recorded. Before touching a formula, identify whether the random variable counts successes inside a fixed n or counts trials until the first success.
Interpretation. Under the historical geometric model, the expected trial number of the first success is 1/p≈11.11. That waiting-time interpretation is fundamentally different from a binomial count, even though both models use two outcome categories and a constant p.
Decision case 27: Warehouse Audit
Design. The process is to inspect 35 pallets and count label mismatches with success probability 0.05 on each independent trial.
Model decision. This is binomial because the audit has a predetermined number of pallets. Before touching a formula, identify whether the random variable counts successes inside a fixed n or counts trials until the first success.
Interpretation. If the independence and constant-p assumptions are defensible, the long-run count has mean 1.75 and standard deviation 1.29. Those summaries describe repeated 35-trial batches; they do not describe the probability of a single trial.
Decision case 28: Network Handshake
Design. The process is to retry a handshake until the first successful connection with success probability 0.52 on each independent trial.
Model decision. This is geometric because the procedure ends when the first connection succeeds. Before touching a formula, identify whether the random variable counts successes inside a fixed n or counts trials until the first success.
Interpretation. Under the historical geometric model, the expected trial number of the first success is 1/p≈1.92. That waiting-time interpretation is fundamentally different from a binomial count, even though both models use two outcome categories and a constant p.
Decision case 29: Training Quiz
Design. The process is to answer 24 independent true-false items and count correct responses with success probability 0.50 on each independent trial.
Model decision. This is binomial because the total number of items is fixed in advance. Before touching a formula, identify whether the random variable counts successes inside a fixed n or counts trials until the first success.
Interpretation. If the independence and constant-p assumptions are defensible, the long-run count has mean 12.00 and standard deviation 2.45. Those summaries describe repeated 24-trial batches; they do not describe the probability of a single trial.
Decision case 30: Donation Kiosk
Design. The process is to observe transactions until the first optional donation occurs with success probability 0.13 on each independent trial.
Model decision. This is geometric because the random variable is a waiting time to first donation. Before touching a formula, identify whether the random variable counts successes inside a fixed n or counts trials until the first success.
Interpretation. Under the historical geometric model, the expected trial number of the first success is 1/p≈7.69. That waiting-time interpretation is fundamentally different from a binomial count, even though both models use two outcome categories and a constant p.
Decision case 31: Lab Batch
Design. The process is to test 40 specimens and count those above a threshold with success probability 0.18 on each independent trial.
Model decision. This is binomial because exactly 40 specimens are tested. Before touching a formula, identify whether the random variable counts successes inside a fixed n or counts trials until the first success.
Interpretation. If the independence and constant-p assumptions are defensible, the long-run count has mean 7.20 and standard deviation 2.43. Those summaries describe repeated 40-trial batches; they do not describe the probability of a single trial.
Decision case 32: Repair Attempts
Design. The process is to attempt repairs until the first machine passes validation with success probability 0.29 on each independent trial.
Model decision. This is geometric because the number of attempts is determined by the first success. Before touching a formula, identify whether the random variable counts successes inside a fixed n or counts trials until the first success.
Interpretation. Under the historical geometric model, the expected trial number of the first success is 1/p≈3.45. That waiting-time interpretation is fundamentally different from a binomial count, even though both models use two outcome categories and a constant p.
Decision case 33: Course Registration
Design. The process is to sample 50 students and count those enrolled full time with success probability 0.74 on each independent trial.
Model decision. This is binomial because the sample size is fixed at 50. Before touching a formula, identify whether the random variable counts successes inside a fixed n or counts trials until the first success.
Interpretation. If the independence and constant-p assumptions are defensible, the long-run count has mean 37.00 and standard deviation 3.10. Those summaries describe repeated 50-trial batches; they do not describe the probability of a single trial.
Decision case 34: Weather Alert
Design. The process is to observe storm cells until the first produces hail with success probability 0.08 on each independent trial.
Model decision. This is geometric because observation continues until first success. Before touching a formula, identify whether the random variable counts successes inside a fixed n or counts trials until the first success.
Interpretation. Under the historical geometric model, the expected trial number of the first success is 1/p≈12.50. That waiting-time interpretation is fundamentally different from a binomial count, even though both models use two outcome categories and a constant p.
Decision case 35: Inventory Scan
Design. The process is to scan 22 items and count barcode failures with success probability 0.04 on each independent trial.
Model decision. This is binomial because the count is measured over a fixed set of 22 scans. Before touching a formula, identify whether the random variable counts successes inside a fixed n or counts trials until the first success.
Interpretation. If the independence and constant-p assumptions are defensible, the long-run count has mean 0.88 and standard deviation 0.92. Those summaries describe repeated 22-trial batches; they do not describe the probability of a single trial.
Decision case 36: Help-Desk Queue
Design. The process is to process tickets until the first password-reset ticket appears with success probability 0.21 on each independent trial.
Model decision. This is geometric because the queue position of the first success is the response. Before touching a formula, identify whether the random variable counts successes inside a fixed n or counts trials until the first success.
Interpretation. Under the historical geometric model, the expected trial number of the first success is 1/p≈4.76. That waiting-time interpretation is fundamentally different from a binomial count, even though both models use two outcome categories and a constant p.
Decision case 37: Inspection Route
Design. The process is to visit 30 sites and count sites needing follow-up with success probability 0.12 on each independent trial.
Model decision. This is binomial because all 30 sites are visited before the count is finalized. Before touching a formula, identify whether the random variable counts successes inside a fixed n or counts trials until the first success.
Interpretation. If the independence and constant-p assumptions are defensible, the long-run count has mean 3.60 and standard deviation 1.78. Those summaries describe repeated 30-trial batches; they do not describe the probability of a single trial.
Decision case 38: Game Challenge
Design. The process is to attempt levels until the first flawless completion with success probability 0.17 on each independent trial.
Model decision. This is geometric because the stopping rule is first success. Before touching a formula, identify whether the random variable counts successes inside a fixed n or counts trials until the first success.
Interpretation. Under the historical geometric model, the expected trial number of the first success is 1/p≈5.88. That waiting-time interpretation is fundamentally different from a binomial count, even though both models use two outcome categories and a constant p.
Decision case 39: Membership Renewal
Design. The process is to contact 26 members and count renewals with success probability 0.61 on each independent trial.
Model decision. This is binomial because renewals are counted among a fixed 26 contacts. Before touching a formula, identify whether the random variable counts successes inside a fixed n or counts trials until the first success.
Interpretation. If the independence and constant-p assumptions are defensible, the long-run count has mean 15.86 and standard deviation 2.49. Those summaries describe repeated 26-trial batches; they do not describe the probability of a single trial.
Decision case 40: Backup Recovery
Design. The process is to retry backups until the first verified restore with success probability 0.44 on each independent trial.
Model decision. This is geometric because the first verified restore ends the sequence. Before touching a formula, identify whether the random variable counts successes inside a fixed n or counts trials until the first success.
Interpretation. Under the historical geometric model, the expected trial number of the first success is 1/p≈2.27. That waiting-time interpretation is fundamentally different from a binomial count, even though both models use two outcome categories and a constant p.
Decision case 41: Product Trial
Design. The process is to give exactly 48 customers a sample and count purchases with success probability 0.27 on each independent trial.
Model decision. This is binomial because there are 48 predetermined trials. Before touching a formula, identify whether the random variable counts successes inside a fixed n or counts trials until the first success.
Interpretation. If the independence and constant-p assumptions are defensible, the long-run count has mean 12.96 and standard deviation 3.08. Those summaries describe repeated 48-trial batches; they do not describe the probability of a single trial.
Decision case 42: Factory Reset
Design. The process is to repeat reset attempts until the first successful diagnostic with success probability 0.34 on each independent trial.
Model decision. This is geometric because the attempt number of first success is random. Before touching a formula, identify whether the random variable counts successes inside a fixed n or counts trials until the first success.
Interpretation. Under the historical geometric model, the expected trial number of the first success is 1/p≈2.94. That waiting-time interpretation is fundamentally different from a binomial count, even though both models use two outcome categories and a constant p.
Decision case 43: Clinic Reminders
Design. The process is to send 36 reminders and count patients who confirm with success probability 0.68 on each independent trial.
Model decision. This is binomial because the number of reminders is fixed. Before touching a formula, identify whether the random variable counts successes inside a fixed n or counts trials until the first success.
Interpretation. If the independence and constant-p assumptions are defensible, the long-run count has mean 24.48 and standard deviation 2.80. Those summaries describe repeated 36-trial batches; they do not describe the probability of a single trial.
Decision case 44: Document Search
Design. The process is to open files until the first contains the requested clause with success probability 0.11 on each independent trial.
Model decision. This is geometric because search ends on first success. Before touching a formula, identify whether the random variable counts successes inside a fixed n or counts trials until the first success.
Interpretation. Under the historical geometric model, the expected trial number of the first success is 1/p≈9.09. That waiting-time interpretation is fundamentally different from a binomial count, even though both models use two outcome categories and a constant p.
Decision case 45: Road Survey
Design. The process is to observe 55 vehicles and count electric vehicles with success probability 0.19 on each independent trial.
Model decision. This is binomial because the observation window fixes n=55 vehicles. Before touching a formula, identify whether the random variable counts successes inside a fixed n or counts trials until the first success.
Interpretation. If the independence and constant-p assumptions are defensible, the long-run count has mean 10.45 and standard deviation 2.91. Those summaries describe repeated 55-trial batches; they do not describe the probability of a single trial.
Decision case 46: Software Test
Design. The process is to rerun an automated test until the first pass with success probability 0.63 on each independent trial.
Model decision. This is geometric because the first pass determines stopping. Before touching a formula, identify whether the random variable counts successes inside a fixed n or counts trials until the first success.
Interpretation. Under the historical geometric model, the expected trial number of the first success is 1/p≈1.59. That waiting-time interpretation is fundamentally different from a binomial count, even though both models use two outcome categories and a constant p.
Decision case 47: Class Survey
Design. The process is to ask 32 students and count those preferring option A with success probability 0.46 on each independent trial.
Model decision. This is binomial because the survey includes exactly 32 students. Before touching a formula, identify whether the random variable counts successes inside a fixed n or counts trials until the first success.
Interpretation. If the independence and constant-p assumptions are defensible, the long-run count has mean 14.72 and standard deviation 2.82. Those summaries describe repeated 32-trial batches; they do not describe the probability of a single trial.
Decision case 48: Medical Device Reboot
Design. The process is to reboot until the first stable reading with success probability 0.38 on each independent trial.
Model decision. This is geometric because the waiting time to the first stable reading is recorded. Before touching a formula, identify whether the random variable counts successes inside a fixed n or counts trials until the first success.
Interpretation. Under the historical geometric model, the expected trial number of the first success is 1/p≈2.63. That waiting-time interpretation is fundamentally different from a binomial count, even though both models use two outcome categories and a constant p.
Decision case 49: Shipping Audit
Design. The process is to check 42 deliveries and count damaged packages with success probability 0.09 on each independent trial.
Model decision. This is binomial because the damage count is over 42 fixed deliveries. Before touching a formula, identify whether the random variable counts successes inside a fixed n or counts trials until the first success.
Interpretation. If the independence and constant-p assumptions are defensible, the long-run count has mean 3.78 and standard deviation 1.85. Those summaries describe repeated 42-trial batches; they do not describe the probability of a single trial.
Decision case 50: Security Login
Design. The process is to attempt authentication until the first accepted credential with success probability 0.57 on each independent trial.
Model decision. This is geometric because the process stops at first acceptance. Before touching a formula, identify whether the random variable counts successes inside a fixed n or counts trials until the first success.
Interpretation. Under the historical geometric model, the expected trial number of the first success is 1/p≈1.75. That waiting-time interpretation is fundamentally different from a binomial count, even though both models use two outcome categories and a constant p.
Binomial vs Geometric Distribution: How to Choose: multiple-choice practice
Comparison question 1
A laboratory tests exactly 20 samples and counts contaminated samples. Which model structure fits if tests are independent with constant contamination probability?
Answer: B
The number of tests is fixed at 20 and X counts successes, so the binomial structure fits.
Comparison question 2
A technician repeats a calibration until the first pass. In an older textbook, which historical model describes the attempt number when p is constant?
Answer: C
The trial count is not fixed; X is the attempt of the first success, which is the historical geometric waiting-time setup.
Comparison question 3
A player takes 15 shots, but success probability rises after every make. What is the best conclusion?
Answer: C
A constant success probability is required for both simple Bernoulli models; adaptive p breaks that condition.
Comparison question 4
A survey contacts exactly 40 people and asks for P(X≥6) responses. Which event translation is correct for a binomial X?
Answer: B
At least 6 includes 6, so use the complement of 0 through 5.
Comparison question 5
An archived waiting-time problem asks for P(X>4) before the first success. What event occurs?
Answer: B
For a first-success waiting time, X>4 means no success occurs in trials 1 through 4.
Comparison question 6
Which statement correctly contrasts the means?
Answer: B
The means summarize different random variables: successes in fixed n versus trials until first success.
Comparison question 7
Which model has support 0,1,…,n for a fixed n?
Answer: A
A binomial count cannot exceed n and can equal zero; a first-success waiting time begins at 1 and has no fixed maximum.
Comparison question 8
A quality inspector samples 12 of only 20 items without replacement. Why is a simple binomial model questionable?
Answer: A
Sampling 60% of the population without replacement changes success probabilities materially from draw to draw.
Comparison question 9
For current 2027 AP Statistics preparation, how should geometric distribution be treated?
Answer: B
College Board removed geometric distribution from the revised course effective in 2026–27.
Comparison question 10
A process continues until the second success. Is the simple first-success geometric model appropriate?
Answer: B
The simple geometric model used in older AP materials describes waiting until the first success, not the second.
Comparison question 11
A campaign sends exactly 100 messages and records the number of replies. Which feature matters most for model selection?
Answer: B
A fixed trial count with success count points toward binomial after independence and constant-p checks.
Comparison question 12
Why is “two outcomes” alone insufficient to justify a binomial model?
Answer: B
Binary outcomes are only one condition; the design must also satisfy the fixed-trial and probability assumptions.
Comparison question 13
A researcher samples 25 records from a database of 10,000 without replacement and counts errors. Which statement best supports a binomial approximation?
Answer: A
The small sampling fraction makes the without-replacement dependence weak enough for the usual approximation when the other binomial conditions hold.
Comparison question 14
Which phrase most clearly signals a waiting-time random variable?
Answer: C
Until the first success defines a random stopping time rather than a fixed number of trials.
Comparison question 15
For X~Bin(20,.30), what does E(X)=6 mean?
Answer: B
Expected value is a long-run average count across repeated fixed-size batches.
Comparison question 16
An older geometric model has p=.10. What does 1/p=10 represent?
Answer: B
The historical geometric mean is a long-run average waiting time, not a deterministic stopping point.
Comparison question 17
A process has fixed n=30 but trials influence one another strongly. What should happen before using a binomial probability?
Answer: B
Fixed n is necessary but not sufficient; dependence can invalidate the binomial model.
Comparison question 18
Which event corresponds to “more than 4 successes” for a binomial count?
Answer: B
More than 4 excludes 4, so the smallest included count is 5.
Comparison question 19
Which statement is most accurate about model labels on current AP Statistics pages?
Answer: B
Clear status labeling prevents older practice from distorting preparation for the revised course.
Comparison question 20
A process stops at the first success, but p differs on weekends and weekdays. What is the key issue?
Answer: B
A first-success structure alone is not enough; the simple historical geometric model also assumes the same success probability each trial.
Binomial vs Geometric Distribution: How to Choose: free-response practice
Comparison FRQ 1
A school gives exactly 30 independent practice items, each with success probability .60. Define X as the number correct. Explain why X is binomial, give its mean and SD, and explain why this is not a first-success waiting-time variable.
Model response
n is fixed at 30, outcomes are success/failure, trials are independent, and p=.60 stays constant. Thus X~Bin(30,.60), with mean 18 and SD √[30(.60)(.40)]≈2.683. The variable counts successes across all 30 trials; it does not stop when the first correct answer occurs.
Comparison FRQ 2
An archived exercise repeats attempts until the first success with p=.25. Explain the historical model, calculate the probability that more than 3 attempts are needed, and state its current-course status.
Model response
Historically this is geometric because X is the trial number of the first success. More than 3 attempts means three initial failures, so P(X>3)=.75^3=.421875. Geometric distribution was removed from the revised AP Statistics tested course, so this should be treated as legacy enrichment.
Comparison FRQ 3
A factory inspects 25 items, but items come in groups from the same machine and outcomes within a group are strongly correlated. Explain why a binomial calculation could mislead even though n is fixed.
Model response
The fixed n and binary outcome conditions are not enough. Strong within-machine dependence violates the independent-trial assumption, so the usual binomial probability formula can give misleading probabilities and variability.
Comparison FRQ 4
A tutoring app increases task difficulty after every success. A student records the trial of the first success. Explain why the historical geometric model may fail.
Model response
Although the response is a first-success waiting time, the app changes task difficulty and therefore can change p from trial to trial. The constant-p condition is not satisfied, so the simple geometric formula is not justified.
Comparison FRQ 5
A company sends exactly 50 independent offers with acceptance probability .08. Find the mean and standard deviation of the acceptance count and explain the model choice.
Model response
The count is binomial because n=50 is fixed, outcomes can be classified as accept/not accept, trials are assumed independent, and p=.08 is constant. The mean is np=4 and the SD is sqrt[50(.08)(.92)]≈1.918.
Comparison FRQ 6
An archived problem asks for the probability the first success occurs on trial 4 when p=.30. Show the historical calculation and explain why it is not a current-core priority.
Model response
The first three trials must fail and trial 4 must succeed, giving (.70)^3(.30)=.1029. This is a historical geometric waiting-time calculation. Geometric distribution was removed from the revised tested course, so it belongs in legacy comparison rather than core 2027 preparation.
Comparison FRQ 7
A researcher plans 20 trials but stops early after a success and records only the stopping trial. Explain why the originally planned maximum of 20 does not automatically make the observed variable binomial.
Model response
The observed random variable is a stopping time, not the number of successes among all 20 planned trials. Model choice follows the variable actually recorded and the stopping rule, not merely the maximum number of trials that could have occurred.
Comparison FRQ 8
A sample of 15 is drawn without replacement from a population of 30 and X counts successes. Identify the main threat to a binomial approximation.
Model response
Although n is fixed and X is a success count, the sample is half the population. Without replacement, each draw materially changes the composition and success probability, so independence and constant-p approximations are weak.
