Lesson 8 of 8 · 18 min
Modelling with binomial distributions
Solve unfamiliar problems with binomial models, including finding the number of trials needed and two-stage problems, and evaluate whether the model's assumptions are reasonable.
What you will be able to do
- Set up a binomial model from a context: define , identify and , and justify the conditions.
- Find the smallest for which a probability such as reaches a target.
- Combine binomial models in two-stage problems.
- Evaluate the assumptions of a binomial model in context.
A nursery sells punnets of 10 seedlings. Past records show 5% of seedlings fail. A customer is happy if at most one seedling in a punnet fails. A café orders 8 punnets. What is the chance that at least 7 of them make the customer happy? Questions like this need two binomial models, one inside the other, and a clear plan.
A plan for binomial modelling
Define the random variable in words
"Let be the number of seedlings in a punnet that fail."
State the distribution and justify it
", assuming seedlings fail independently with the same probability."
Write the probability statement
"Happy" means , so you need .
Calculate
By the formula or with technology, rounding only at the end.
Answer in context and check reasonableness
A sentence that answers the question, and a quick sense check: is the probability between 0 and 1, and about the size you expected?
The number of failed seedlings in a punnet, . Most of the probability is at 0 and 1.
Worked example
Seedling punnets, a two-stage binomial model
A nursery sells punnets of 10 seedlings. Each seedling fails to grow with probability 0.05, independently of the others. A customer is happy with a punnet if at most one of its seedlings fails.
(a) Determine the probability that a customer is happy with a punnet, correct to 4 decimal places.
(b) A café buys 8 punnets. Determine the probability that the café is happy with at least 7 of them, correct to 4 decimal places.
(c) State one assumption needed in (b) and explain whether it is reasonable.
Step 1: Model one punnet
Let be the number of failed seedlings in a punnet: .
Step 2: Find the probability of a happy punnet
Step 3: Model the 8 punnets
Let be the number of happy punnets: with (kept unrounded in the calculator).
Each punnet is now a Bernoulli trial whose "success" is "happy", with the probability from part (a).
Step 4: Find "at least 7"
Step 5: Evaluate an assumption
The model assumes punnets are independent with the same . Punnets from the same batch share soil, seed stock and watering, so failures may cluster and may differ between batches. If so, the true probability could be lower than 0.8533.
Try it yourself first, then check the working one step at a time.
Answer
(a) . (b) . (c) Punnets are assumed independent with the same probability of being satisfactory; this is doubtful if they come from one batch, because failures could cluster.
Method
Solve the inner problem: a probability for one group.
Use that probability as in a second binomial model for the number of groups.
Keep unrounded values in the calculator between stages.
Evaluate independence and constant in context.
Watch out for
Using 0.95 as the probability that a punnet is satisfactory.
Rounding to 0.91 before the second stage.
Calculating only for "at least 7".
Each punnet makes the customer happy with probability 0.9139 (from the example). Find the probability that all 8 punnets make the customer happy, correct to 4 decimal places.
Another common question turns the problem around: how many trials do you need? "At least one" is the complement of "none", and , so the condition becomes an inequality in that you solve with logarithms or by trial with a calculator.
Worked example
How many bottles for a likely win?
In a promotion, 1 in 10 bottle caps wins a prize, independently. Determine the smallest number of bottles you must buy so that the probability of at least one winning cap is at least 0.95.
Step 1: Define the model
Let be the number of winning caps in bottles. .
Step 2: Use the complement
"At least one" has many terms; "none" has one.
Step 3: Write and rearrange the inequality
Step 4: Solve with logarithms
Dividing by , which is negative, reverses the inequality.
Step 5: Round up and check
. Check: , but .
Try it yourself first, then check the working one step at a time.
Answer
You must buy at least 29 bottles.
Method
Write .
Set up the inequality and solve for with logarithms (or a table of values).
Round up to a whole number, then check and .
Watch out for
Forgetting to reverse the inequality when dividing by a negative logarithm.
Rounding 28.43 down to 28, which doesn't reach 0.95.
A rare plant species is found in 15% of quadrats surveyed in a rainforest. Assuming quadrats are independent, what is the smallest number of quadrats an ecologist should survey to have a probability of at least 0.9 of finding the plant at least once?
Common mistake
Rounding n the wrong way
The number of trials must be a whole number that meets the condition, so always round up, then check the value and the one below it. A probability of 0.897 does not meet "at least 0.9".
A model is only as good as its assumptions. In context, ask:
- Is constant? A goal kicker tires late in a game; a machine drifts out of tune.
- Are the trials independent? People in the same household tend to get sick together; passengers booked on one ticket turn up together.
- Is fixed? "Keep trying until you succeed" doesn't have a fixed number of trials.
- Is the sample small compared with the population? If not, drawing without replacement changes .
A school models the number of students absent from a class of 25 on a given day as . Which is the strongest reason the model may be unreliable during flu season?
Exam tip
Evaluate in context
When asked to evaluate or comment on a model, name the condition, say how this context could break it, and say what effect that has on the answer (too high or too low). Generic statements like "the model may not be accurate" earn nothing.
Summary
Define , state with reasons, write the probability statement, calculate, answer in context.
"At least one" is . To find the smallest , solve target with logarithms and round up.
In two-stage problems, a probability from one binomial model becomes in the next.
Evaluate a model by checking constant , independence, fixed and the sample size, in context.
Finished the lesson?
Try the 3 quick checks above and the lesson completes itself, or mark it complete now.