Common Errors in Binomial and Geometric Calculations

Why These Calculations Are Easy to Misread

Binomial and geometric calculations use compact formulas, but the symbols only give the right answer when they represent the event in the question. A misplaced exponent can change how many trials are counted. A missing combination factor can count only one possible sequence when many sequences meet the event. A complement can quietly change an inclusive endpoint into a strict one.

A useful habit is to pause before calculating: say what counts as a success, write the event in words or symbols, and identify whether the question concerns a fixed number of trials or the wait until a first success. Then check that the formula counts precisely that event.

Definition: A binomial model counts successes in a fixed number \(n\) of independent trials, each with the same success probability \(p\). The random variable \(X\) is the number of successes. A geometric model records \(T\), the trial number on which the first success occurs, under independent trials with the same success probability \(p\). In either model, outcomes must be classifiable as success or failure according to a clearly defined rule.

The binomial model requires a fixed \(n\), independent trials, the same \(p\) on every trial, and two outcome categories for the count. The geometric model requires repeated independent trials with constant \(p\), continued until a first success. If these assumptions do not fit the situation, correcting arithmetic will not make the model appropriate.

Begin by Defining Success

The word success is a mathematical label for the outcome being counted or awaited; it does not mean that the outcome is desirable. If the question asks for the number of defective items, define “defective” as success. Then \(p\) is the probability an item is defective. Defining “not defective” as success instead would also be possible, but the probability parameter and event would need to change accordingly.

An unclear definition often causes a calculation to use \(p\) where \(1-p\) belongs. Before using a formula, state in context what one success is and what its probability is. For an exact binomial event, also state how many successes are required. For a geometric event, state what first success the waiting time ends at.

Worked Example: Counting Exactly Two Defective Items

Suppose, hypothetically, that five items are inspected independently and each has probability \(0.20\) of being defective. Find the probability that exactly two are defective. Define success as “the inspected item is defective,” so \(p=0.20\), and let \(X\) be the number of defective items among the five.

The binomial assumptions match the setup: there are \(n=5\) fixed inspections, each inspection has two relevant outcomes (defective or not defective), the probability of a defective item is the same for each, and the inspections are stipulated to be independent.

One particular sequence with exactly two defective items, such as defective, defective, not defective, not defective, not defective, has probability \( (0.20)^2(0.80)^3 \). The factors are \(0.20\) for each of its two defective outcomes and \(0.80\) for each of its three nondefective outcomes. Thus,

$$ (0.20)^2(0.80)^3 =0.04(0.512) =0.02048. $$

But that is the probability of one specified order, not of exactly two defective items in any order. There are \(\binom{5}{2}=10\) ways to choose which two of the five items are defective. Each qualifying sequence has the same probability, so the exact-count probability is

P(X=2) =\binom{5}{2}(0.20)^2(0.80)^3 =10(0.02048) =0.2048.

The probability that exactly two of the five items are defective is \(0.2048\), or \(20.48\%\). Multiplying by 10 is necessary because the event includes 10 different sequences. Omitting that factor answers a narrower question: the probability of one particular sequence.

Why the Combination Factor Is Needed

For an exact binomial count \(X=k\), each sequence with \(k\) successes and \(n-k\) failures has probability \(p^k(1-p)^{n-k}\). The number of different sequences with \(k\) successes is \(\binom{n}{k}\), because choosing the \(k\) trial positions for successes determines the sequence. The sequences are mutually exclusive, so their probabilities add.

P(X=k)=\binom{n}{k}p^k(1-p)^{n-k}.

This gives a short justification of the exact-count formula: there are \(\binom{n}{k}\) equally probable qualifying sequences, and each has probability \(p^k(1-p)^{n-k}\). Multiplication counts all of them. The combination factor is not needed when the event specifies one particular ordered sequence; it is needed when the event includes every arrangement with the stated number of successes.

A useful check is to ask: Does the event care about the order? “Success on trials 1 and 3, and failure on trial 2” specifies one sequence. “Exactly two successes in three trials” includes three sequences, so the count is \(\binom{3}{2}\).

Match Geometric Exponents to the Waiting Event

For a geometric random variable \(T\), the exponent counts failures before the first success. To have the first success on trial \(k\), the first \(k-1\) trials must fail, followed by a success on trial \(k\). Therefore,

P(T=k)=(1-p)^{k-1}p,\qquad k=1,2,3,\ldots.

The exponent is \(k-1\), not \(k\), because success occurs on trial \(k\); only the preceding trials are failures. Writing down the first few cases is an effective check: \(P(T=1)=p\), \(P(T=2)=(1-p)p\), and \(P(T=3)=(1-p)^2p\).

Worked Example: The First Success on Trial Three

Suppose a hypothetical device test has independent repeated trials, each with probability \(p=0.25\) of detecting a fault. Let \(T\) be the trial number of the first detection. Find the probability that the first detection occurs on trial 3.

The geometric model is appropriate under the stated assumptions: each trial has the same probability \(0.25\) of detection, trials are independent, and testing continues until the first detection. For \(T=3\), trials 1 and 2 must be failures, and trial 3 must be a success. Hence,

P(T=3) =(1-0.25)^{3-1}(0.25) =(0.75)^2(0.25) =0.5625(0.25) =0.140625.

Thus, the probability of a first detection on trial 3 is \(0.1406\), or about \(14.06\%\). The incorrect expression \((0.75)^3(0.25)\) would require three failures followed by a success, which describes a first detection on trial 4, not trial 3.

Inclusive Endpoints and Complements

Words such as at least, more than, at most, and fewer than determine which endpoint is included. For a geometric waiting time, \(T\leq m\) means a first success occurs by trial \(m\); \(T<m\) stops at trial \(m-1\). Likewise, \(T\geq m\) includes trial \(m\), while \(T>m\) excludes it.

Complements must preserve these distinctions. For example, the complement of \(T\leq m\) is \(T>m\), not \(T\geq m\). The complement of \(T\geq m\) is \(T<m\). For geometric trials, surviving past trial \(m\) means the first \(m\) trials were all failures, giving \(P(T>m)=(1-p)^m\). Reaching trial \(m\) or later requires the first \(m-1\) trials to be failures, giving \(P(T\geq m)=(1-p)^{m-1}\).

Worked Example: “By Trial Three” Is Not “After Trial Three”

For the hypothetical geometric setting above, calculate (a) the probability of a first detection by trial 3 and (b) the probability that the first detection occurs on trial 3 or later. The same geometric assumptions apply: independent trials with constant detection probability \(0.25\), continued until the first detection.

(a) “By trial 3” is \(T\leq 3\). Its complement is \(T>3\), meaning that all of the first three trials fail. Therefore,

P(T\leq 3) =1-P(T>3) =1-(0.75)^3 =1-0.421875 =0.578125.

The probability of a first detection by trial 3 is \(0.5781\), or about \(57.81\%\). Using \(1-(0.75)^2\) would calculate the probability of success by trial 2, not by trial 3.

(b) “On trial 3 or later” is \(T\geq 3\), so trials 1 and 2 must both fail. Trial 3 may be either outcome. Thus,

P(T\geq 3) =(0.75)^2 =0.5625.

A common endpoint error is to write \(1-P(T\leq 3)\) for this event. That expression is \(P(T>3)=0.421875\), which requires the first three trials to fail and excludes a first detection on trial 3. Since \(T\geq 3\) includes trial 3, its probability is instead \(0.5625\).

Use Complements Only When They Simplify the Event

A complement is helpful when it replaces many cases with one simple event, but it is not automatically the easiest route. For binomial counts, “at least one success” is naturally calculated as one minus “zero successes.” “At most one success,” by contrast, is not the complement of “zero successes”; its complement is “two or more successes.” Naming the complement explicitly prevents accidentally leaving out or adding a boundary case.

Worked Example: At Least One Success in Four Trials

Suppose, hypothetically, that four independent trials each have probability \(0.30\) of success. Find the probability of at least one success. Let \(X\) count successes, so the event is \(X\geq 1\). The binomial assumptions hold by the setup: there are four fixed trials, two outcomes per trial, constant success probability, and independence.

The complement of \(X\geq1\) is \(X=0\), which means all four trials fail. Each failure has probability \(1-0.30=0.70\), so

P(X\geq1) =1-P(X=0) =1-(0.70)^4 =1-0.2401 =0.7599.

The probability of at least one success is \(0.7599\), or \(75.99\%\). The complement includes exactly one outcome pattern—four failures—so no combination factor is needed. A frequent mistake is to subtract \(P(X=1)\) from 1; that leaves both zero successes and two or more successes, rather than the desired event.

A Practical Error-Checking Routine

When a result seems surprising, check the setup and the event before repeating the arithmetic. The following sequence catches most errors in binomial and geometric calculations.

Name success and \(p\).
Write what outcome is called a success and state its probability. If the question counts failures, decide whether to redefine success or use failure probability \(1-p\), and remain consistent.

Identify the random variable and event.
Use \(X\) for a count in a fixed \(n\) trials and \(T\) for a first-success trial number. Translate words like “exactly,” “by,” and “at least” into an event before choosing a formula.

Check what the formula counts.
For an exact binomial count, decide whether the event is one ordered sequence or all arrangements with \(k\) successes. For a geometric probability, verify that the exponent counts failures before success.

Verify a complement and its endpoint.
Say the complement in words. Confirm that it includes exactly the outcomes omitted by the original event, especially at an inclusive boundary.

Common Mistakes

Using an undefined success. The correction is to identify the counted or awaited outcome first, then attach \(p\) to that outcome.
Forgetting the combination factor. The probability of one sequence is not the probability of every arrangement with the same success count. Multiply by \(\binom{n}{k}\) when all arrangements qualify.
Using \(k\) failures for a first success on trial \(k\). There are only \(k-1\) failures before that success, so the geometric exponent is \(k-1\).
Replacing “at least” with “more than.” \(T\geq m\) includes trial \(m\); \(T>m\) does not. Write the event symbolically before taking a complement.
Taking a complement that is too narrow or too broad. For “at least one,” the complement is exactly zero. For “at most one,” it is two or more—not exactly one.

Key takeaway: A correct calculation must count the right outcomes exactly once. Define success, translate the event, match the exponent and combination count to that event, and check that any complement has the right endpoint.

Check Your Understanding

For each question, focus on the event being counted before calculating.

In five independent trials with success probability \(0.10\), what does \((0.10)^2(0.90)^3\) represent, and what extra factor is needed for exactly two successes?
A geometric waiting time is \(T\). Explain why \(P(T=4)=(1-p)^3p\), rather than \((1-p)^4p\).
State the complement of \(T\leq 6\) and describe it in words.
For a binomial count \(X\), what is the complement of \(X\geq1\)? Why is this complement especially simple to calculate?
Explain the difference between “first success on trial 5 or later” and “first success after trial 5.”