Read the Condition Before Choosing a Denominator
In Conditional Probability with Dice and Cards, you restricted the sample space to the event named by the condition. The same principle applies when a question gives counts in a table or describes groups in words: the condition identifies the reference group, and the denominator must come from that group.
Many conditional-probability errors are not arithmetic errors. They happen because a calculation answers a different question from the one asked. A student may use the full group instead of the group specified by “given that,” or may reverse the target event and the condition. Before dividing, translate the question into \(P(A\mid B)\): \(A\) is the event whose probability is requested, and \(B\) is the condition.
For counts, this means dividing the number in both \(A\) and \(B\) by the number in \(B\). For probabilities, divide \(P(A\cap B)\) by \(P(B)\), as in The Conditional Probability Formula. The numerator must be contained in the denominator group: every outcome counted in \(A\cap B\) must also be counted in \(B\).
A useful reading habit is to start with the words after “given that.” Those words identify \(B\). Then ask what proportion of that group satisfies \(A\). This simple order helps prevent both the wrong-denominator error and the condition-swapping error.
Worked Example: Workshop Participation and Passing
Worked Example: Workshop Participation and Passing
A school’s illustrative records classify 120 students by whether they attended a study workshop and whether they passed a course assessment. Of the 60 workshop attendees, 42 passed. Of the 60 students who did not attend, 28 passed. Among all 120 students, what is the probability that a randomly selected student passed, given that the student attended the workshop?
| Passed | Did not pass | Total | |
|---|---|---|---|
| Attended workshop | 42 | 18 | 60 |
| Did not attend | 28 | 32 | 60 |
| Total | 70 | 50 | 120 |
State: Let \(P\) mean that a student passed and \(W\) mean that a student attended the workshop. The question asks for \(P(P\mid W)\).
Plan: The condition is workshop attendance, so use the 60 workshop attendees as the reference group. Within that group, count the students who passed. The denominator must be 60, not the 120 students overall.
Do: There are 42 students in both the “attended” row and the “passed” column.
Conclude: Among the students who attended the workshop, the probability of passing was \(0.700\), or 70%. This describes the proportion of workshop attendees who passed; it does not describe the proportion of all students who passed.
A common wrong calculation is \(42/120=0.350\). That is the joint probability \(P(P\cap W)\): the chance that a randomly selected student from the entire group both attended and passed. Its denominator includes students who did not attend, so it does not answer the conditional question. The table helps diagnose the mismatch: the question says “given that the student attended,” but 120 is the grand total rather than the workshop row total.
Swapping the Condition Changes the Question
The same two events can form two different conditional probabilities. \(P(A\mid B)\) asks for the proportion of the \(B\) group that has characteristic \(A\). \(P(B\mid A)\) asks for the proportion of the \(A\) group that has characteristic \(B\). The overlap is the same, but the reference groups—and therefore the denominators—are different.
This is the issue discussed in Why \(P(A\mid B)\) Is Not \(P(B\mid A)\). Do not assume that reversing the event names leaves a probability unchanged. Instead, write down the event after “given that,” then identify its total before calculating.
Worked Example: Rainy Days and Late Buses
Worked Example: Rainy Days and Late Buses
In a fictional set of 100 recorded bus trips, 30 trips occurred on rainy days. The bus was late on 12 rainy trips and 14 trips that were not rainy. What are (a) the probability a trip was late given that it was rainy and (b) the probability a trip was rainy given that the bus was late?
State: Let \(L\) mean that a bus trip was late, and \(R\) mean that it occurred on a rainy day. Part (a) asks for \(P(L\mid R)\). Part (b) asks for \(P(R\mid L)\).
Plan: For part (a), the condition \(R\) selects the 30 rainy trips. For part (b), the condition \(L\) selects all late trips. There were 12 late rainy trips and 14 late non-rainy trips, so there were \(12+14=26\) late trips in total. Both questions have the same overlap count, 12, but different condition groups.
Do: For part (a), divide the late rainy trips by all rainy trips. For part (b), divide the late rainy trips by all late trips.
Conclude: On rainy trips, the probability that a bus was late was 0.400, or 40%. Among late trips, the probability that a trip occurred on a rainy day was about 0.4615, or 46.15%. These probabilities differ because they describe proportions within different groups.
Using 100 as the denominator in either part would instead give the joint probability that a randomly selected trip was both rainy and late: \(12/100=0.120\). Using 30 for both answers would also be an error: 30 is the rainy-trip total, but part (b) is restricted to late trips. A quick check is to point to the condition in each part and confirm that its count is the denominator.
Check That the Numerator Belongs to the Condition Group
A reliable way to catch mistakes is to check the relationship between the numerator and denominator before dividing. In a count table, the overlap count must be inside the row or column used as the condition. In probability notation, \(A\cap B\) is part of \(B\), so \(P(A\cap B)\) cannot be greater than \(P(B)\). A conditional probability must be between 0 and 1.
The quotient can also be checked against its meaning. If you calculate \(P(A\mid B)\), explain the result as a proportion of the \(B\) group. If your sentence instead describes all outcomes, you may have used a joint probability. If it describes the \(A\) group, you may have reversed the condition.
Worked Example: Irrigation and Plant Survival
Worked Example: Irrigation and Plant Survival
For an invented garden record, 80 planted plots are classified by whether they received automated irrigation and whether the plants survived the season. In 45 irrigated plots, plants survived in 36. In 35 plots without automated irrigation, plants survived in 21. What is the probability that plants survived, given that a plot received automated irrigation?
| Plants survived | Plants did not survive | Total | |
|---|---|---|---|
| Automated irrigation | 36 | 9 | 45 |
| No automated irrigation | 21 | 14 | 35 |
| Total | 57 | 23 | 80 |
State: Let \(S\) mean that the plants survived and \(I\) mean that a plot received automated irrigation. We want \(P(S\mid I)\).
Plan: Irrigation is the condition, so the reference group is the 45 irrigated plots. Within those plots, 36 had plants that survived. The condition has a positive count, so the conditional probability is defined.
Do: Divide the number of irrigated plots where the plants survived by the number of irrigated plots.
Conclude: In these illustrative records, the probability that plants survived, given that a plot received automated irrigation, is 0.800, or 80%. It is the survival proportion among irrigated plots, not among all 80 plots.
As a check, the joint probability that a randomly selected plot was irrigated and had surviving plants is \(36/80=0.450\), while the probability a plot was irrigated is \(45/80=0.5625\). Applying the formula gives
This check reaches the same answer as the table calculation. By contrast, \(36/80=0.450\) alone would answer a different question because it uses the full set of plots as its reference group.
Other Errors to Catch
- Using the overlap count as the answer. The number in both events is a count, not a conditional probability. Divide it by the total in the condition group.
- Using the grand total automatically. The grand total is the denominator for a probability from the full group, such as a joint probability. It is not automatically the denominator for a conditional probability.
- Swapping the target and condition. Read the wording carefully and write \(P(\text{target}\mid\text{condition})\). The event after the bar sets the reference group.
- Using a condition group with no outcomes. If \(P(B)=0\), \(P(A\mid B)\) is undefined because there are no outcomes in the reference group. It is not automatically zero.
- Giving only a decimal. A decimal without a reference group can hide a denominator error. State what group the probability describes and include units or context when appropriate.
For full credit, make the reference group visible in your work. For a count table, identify the condition row or column, show the overlap count divided by that group’s total, and interpret the result using “among” or “given that.” For probability values, name the numerator as the joint probability and the denominator as the probability of the condition.
Check Your Understanding
For each situation, identify the condition and explain which group supplies the denominator before calculating.
- In a group of 90 students, 24 take art and play a sport, while 40 play a sport. What is the probability that a student takes art, given that the student plays a sport?
- Using the same counts, what additional total would you need to find the probability that a student plays a sport, given that the student takes art?
- Explain why the joint probability that a bus trip was rainy and late is not the same question as the probability a trip was late given that it was rainy.
- If \(P(B)=0\), is \(P(A\mid B)\) equal to zero, or is it undefined? Explain using the reference-group meaning.
- A student calculates \(P(R\mid L)\) using the total number of rainy trips as the denominator. Identify the likely error and state what group should be the denominator.