Translating Verbal Events Into Binomial Bounds

Why Verbal Bounds Need Care

A binomial model describes a count, so its possible values are integers. A verbal condition such as “fewer than three” therefore identifies particular integer counts, not a continuous range. Translating the words incorrectly by even one endpoint changes the event and can change its probability substantially.

For an actuarial application, the count might represent the number of policy applications with a specified feature, claims meeting a condition, or inspections that detect a defect. In each case, first define what counts as a success and identify the binomial model. Then translate the event into integer values before calculating its probability.

Definition: Suppose $X$ counts successes in $n$ independent trials, each with the same success probability $p$. Then $X\sim\operatorname{Bin}(n,p)$, and its possible values are $0,1,\ldots,n$. For any integer $k$ in this range, $$ P(X=k)=\binom{n}{k}p^k(1-p)^{n-k}. $$ A verbal event involving $X$ is a set of these possible integer values.

This model requires a fixed number $n$ of trials, two outcomes per trial, independent trials, and the same probability $p$ of success on every trial. The translation rules below concern the event wording; they do not make a binomial model appropriate if those assumptions fail.

Translate the Words Before Calculating

For an integer-valued count, the main phrases have direct translations. In the table, $r$ and $k$ are integers, and any event must also be restricted to the support $0,\ldots,n$.

Verbal phrase	Event in symbols	Integer values included
Exactly $k$	$X=k$	Only $k$
Fewer than $r$	$X<r$, or $X\leq r-1$	At most $r-1$
No more than $r$, at most $r$	$X\leq r$	Through $r$, including $r$
More than $k$	$X>k$, or $X\geq k+1$	From $k+1$ upward
At least $k$	$X\geq k$	From $k$ upward, including $k$

The difference between “fewer than” and “no more than” is the endpoint. “Fewer than $r$” excludes $r$; “no more than $r$” includes it. Similarly, “more than $k$” excludes $k$, whereas “at least $k$” includes it.

Once the event has been translated, calculate its probability by adding the exact binomial probabilities for the included values. For example, if the event is $a\leq X\leq b$, its probability is the sum of $P(X=k)$ for each integer $k$ from $a$ through $b$. If a bound falls outside $0,\ldots,n$, only supported values can occur; if no supported values meet the condition, the event is impossible.

Justification: Because $X$ can take only integer values, the integers strictly less than an integer $r$ end at $r-1$. Thus $X<r$ and $X\leq r-1$ describe the same event. Likewise, an integer strictly greater than $k$ must be at least $k+1$, so $X>k$ and $X\geq k+1$ are equivalent. These endpoint shifts follow from the integer support, not from a probability calculation.

Strict Inequalities and Noninteger Cutoffs

Words such as “more than” and “fewer than” can also be applied to noninteger cutoffs. The count remains integer, so convert the cutoff to the nearest relevant integer boundary rather than treating $X$ as continuous. For any real number $a$:

\begin{aligned} X<a &\quad\Longleftrightarrow\quad X\leq \lceil a\rceil-1,\\ X\leq a &\quad\Longleftrightarrow\quad X\leq \lfloor a\rfloor,\\ X>a &\quad\Longleftrightarrow\quad X\geq \lfloor a\rfloor+1,\\ X\geq a &\quad\Longleftrightarrow\quad X\geq \lceil a\rceil. \end{aligned}

Here $\lfloor a\rfloor$ is the greatest integer no larger than $a$, and $\lceil a\rceil$ is the least integer no smaller than $a$. For instance, $X>2.5$ means $X\geq3$, while $X\leq2.5$ means $X\leq2$. When the cutoff is already an integer, strictness still matters: $X>2$ starts at 3, but $X\geq2$ starts at 2.

For a combined condition $a<X<b$, the integer values run from $\lfloor a\rfloor+1$ through $\lceil b\rceil-1$, subject to the binomial support. If $a$ and $b$ are integers, this reduces to $a+1$ through $b-1$. A phrase such as “between $a$ and $b$” can be ambiguous about inclusion, so use an explicit verbal condition or clarify whether endpoints are included.

Worked Example: Fewer Than Two Reported Claims

A hypothetical review examines five independent records. For each record, the probability of a specified claim feature is $0.20$. Let $X$ be the number of records with that feature. Find the probability that fewer than two records have it.

Method and translation. Each record is a Bernoulli trial with success probability $p=0.20$, and the trials are independent, so $X\sim\operatorname{Bin}(5,0.20)$. “Fewer than two” means $X<2$, which for an integer count is $X=0$ or $X=1$.

The failure probability is $1-p=0.80$. Calculate each included exact probability:

\begin{aligned} P(X=0)&=\binom{5}{0}(0.20)^0(0.80)^5=1(1)(0.32768)=0.32768,\\ P(X=1)&=\binom{5}{1}(0.20)^1(0.80)^4=5(0.20)(0.4096)=0.40960. \end{aligned}

Add the probabilities for the disjoint counts:

$$ P(X<2)=P(X=0)+P(X=1)=0.32768+0.40960=0.73728. $$

Rounded to four decimal places, the answer is $0.7373$. The check is that the event includes exactly the two supported counts below 2, namely 0 and 1. The common mistake avoided is including $X=2$, which “fewer than two” explicitly excludes.

Combining Two Strict Conditions

When a statement contains two bounds, translate each side separately, then take the intersection. This prevents the common error of treating a strict endpoint as included. For example, “more than one but fewer than five” requires $X>1$ and $X<5$ at the same time.

Worked Example: More Than One but Fewer Than Five

In a hypothetical set of six independent records, each record has probability $0.40$ of a specified event. Let $X$ count the records with the event. Find the probability that the count is more than one but fewer than five.

Method and translation. Here $X\sim\operatorname{Bin}(6,0.40)$, with failure probability $0.60$. “More than one” means $X\geq2$; “fewer than five” means $X\leq4$. Together, these give $2\leq X\leq4$, so the included counts are 2, 3, and 4.

Calculate each exact probability:

\begin{aligned} P(X=2)&=\binom{6}{2}(0.40)^2(0.60)^4 =15(0.16)(0.1296)=0.31104,\\ P(X=3)&=\binom{6}{3}(0.40)^3(0.60)^3 =20(0.064)(0.216)=0.27648,\\ P(X=4)&=\binom{6}{4}(0.40)^4(0.60)^2 =15(0.0256)(0.36)=0.13824. \end{aligned}

The counts are mutually exclusive, so add their probabilities:

P(2\leq X\leq4)=0.31104+0.27648+0.13824=0.72576.

Rounded to four decimal places, the answer is $0.7258$. As a check, the event includes three counts, all strictly above 1 and strictly below 5. The common mistake avoided is reading “more than one” as “at least one,” which would incorrectly add $P(X=1)$.

Bounds at the Edge of the Support

The binomial count cannot be negative or exceed $n$. Thus a condition such as “at least $n$” includes only the single possible value $n$, while “more than $n$” is impossible. This support check is useful even when the verbal translation itself is correct: it can reveal that some nominal integer values cannot occur.

Worked Example: At Least Six Successful Inspections

A hypothetical batch has eight independent inspected items, each with probability $0.25$ of meeting a specified condition. Let $X$ count the items that meet it. Find the probability that at least six meet the condition.

Method and translation. The model is $X\sim\operatorname{Bin}(8,0.25)$, with failure probability $0.75$. “At least six” means $X\geq6$. Since the support ends at 8, the event consists of $X=6,7,8$.

Using the common denominator $4^8=65536$, compute the three exact probabilities:

\begin{aligned} P(X=6)&=\binom{8}{6}\left(\frac14\right)^6\left(\frac34\right)^2 =\frac{28\cdot9}{65536}=\frac{252}{65536},\\ P(X=7)&=\binom{8}{7}\left(\frac14\right)^7\left(\frac34\right) =\frac{8\cdot3}{65536}=\frac{24}{65536},\\ P(X=8)&=\binom{8}{8}\left(\frac14\right)^8 =\frac{1}{65536}. \end{aligned}

Therefore,

P(X\geq6)=\frac{252+24+1}{65536} =\frac{277}{65536} =0.0042266845703125.

Rounded to four decimal places, the answer is $0.0042$. The support check confirms that no values above 8 need to be considered. The common mistake avoided is interpreting “at least six” as strictly greater than six and omitting $X=6$.

Common Endpoint Errors

A reliable calculation begins with an event statement, not with a probability formula. Write the event in symbols, list its integer values, and only then substitute into the binomial probability mass function. In particular, do not let a familiar phrase obscure whether its endpoint is included.

For the six-record example, an incorrect translation of “more than one but fewer than five” is $1\leq X\leq4$. That mistake treats “more than one” as “at least one.” It produces the wrong calculation:

P(1\leq X\leq4) =P(X=1)+P(X=2)+P(X=3)+P(X=4) =0.186624+0.31104+0.27648+0.13824 =0.912384.

The correction is to exclude 1 because $1$ is not more than $1$. The correct event is $2\leq X\leq4$, and its probability is $0.72576$, or $0.7258$ to four decimal places. The difference between these answers is exactly the incorrectly included probability $P(X=1)=0.186624$.

“Fewer than” is not “no more than.” For an integer $r$, fewer than $r$ ends at $r-1$, while no more than $r$ includes $r$.
“More than” is not “at least.” More than $k$ starts at $k+1$; at least $k$ starts at $k$.
Check support as well as wording. A binomial count cannot be below 0 or above $n$.
Do not round individual terms prematurely. Add the exact or unrounded probabilities, then round the final probability.

Key takeaway: Translate every verbal condition into an integer event before calculating. For strict bounds, exclude the endpoint; for “at least” and “no more than,” include it. Then sum the exact binomial probabilities for precisely the supported values in that event.

Check Your Understanding

For each problem, translate the wording into integer values before finding the probability. Round each final probability to four decimal places.

A hypothetical group has three independent cases, each with probability $0.10$ of a specified outcome. What is the probability of fewer than two outcomes? (Answer: $0.9720$)
For four independent trials with success probability $0.30$, find the probability of no more than one success. (Answer: $0.6517$)
A hypothetical screening process has six independent tests, each positive with probability $0.50$. Find the probability that at least two but no more than four tests are positive. (Answer: $0.7813$)
Seven independent items each have probability $0.25$ of meeting a criterion. Find the probability that fewer than three meet it. (Answer: $0.7564$)
Six independent records each have probability $0.40$ of a specified feature. Find the probability that the count is more than $2.5$ but fewer than $5.2$. (Answer: $0.4516$)

Verbal phrase	Event in symbols	Integer values included
Exactly \(k\)	\(X=k\)	Only \(k\)
Fewer than \(r\)	\(X<r\), or \(X\leq r-1\)	At most \(r-1\)
No more than \(r\), at most \(r\)	\(X\leq r\)	Through \(r\), including \(r\)
More than \(k\)	\(X>k\), or \(X\geq k+1\)	From \(k+1\) upward
At least \(k\)	\(X\geq k\)	From \(k\) upward, including \(k\)

What You'll Learn