Lesson 4 of 8 · 16 min

Probabilities for sample means

Calculate probabilities and quantiles for a sample mean with technology or the 68–95–99.7% rule, compare them with probabilities for one value, and handle totals.

What you will be able to do

  • Calculate P(Xˉ<a)P(\bar{X} < a), P(Xˉ>a)P(\bar{X} > a) and P(a<Xˉ<b)P(a < \bar{X} < b) with technology, and without it when aa and bb are whole numbers of standard errors from μ\mu.
  • Find a value cc with P(Xˉ<c)P(\bar{X} < c) equal to a given probability.
  • Contrast a probability about one value with the same probability about a sample mean.
  • Turn a question about a total of nn values into one about Xˉ\bar{X}.

A flour mill fills bags labelled 1 kg. The filled masses are normal with mean 10051005 g and standard deviation 1212 g. An inspector picks one bag: is it likely to be under 10001000 g? Now she weighs 36 bags and takes their mean: is that likely to be under 10001000 g? The questions sound alike, but the answers are very different.

Shaded: below 1000 g. For one bag (grey) it is about a third of the area. For the mean of 36 bags (orange) it is well under 1%.

Finding a probability for a sample mean

  1. Identify μ\mu, σ\sigma and nn

    Check whether the question is about one value (XX) or a sample mean (Xˉ\bar{X}).

  2. State the distribution of Xˉ\bar{X}, with a reason

    Xˉ∼N(μ,σ2n)\bar{X} \sim N\left(\mu, \frac{\sigma^2}{n}\right): exactly if XX is normal, approximately if nn is large.

  3. Calculate the probability

    With technology, enter the standard deviation σn\frac{\sigma}{\sqrt{n}}. Without it, find how many standard errors the value is from μ\mu and use the 68–95–99.7% rule.

  4. Answer in context

    Round as asked and say what the probability means.

Worked example

One bag or the mean of 36 bags?

  • Simple familiar
  • Technology active

Bags of flour labelled 1 kg have masses that are normally distributed with mean 10051005 g and standard deviation 1212 g.

(a) Find the probability that one randomly chosen bag has mass less than 10001000 g.

(b) Find the probability that the mean mass of a random sample of 3636 bags is less than 10001000 g.

Give your answers correct to 4 decimal places.

4 steps

Try it yourself first, then check the working one step at a time.

Method

  1. One value: use σ\sigma. A sample mean: use σn\frac{\sigma}{\sqrt{n}}.

  2. Write the distribution, then the calculator command with its inputs.

Watch out for

  • Using σ=12\sigma = 12 in part (b), which repeats the answer to part (a).

  • Entering the variance 44 into normCdf instead of the standard deviation 22.

Heights of adult men in a town are normally distributed with mean 170170 cm and standard deviation 88 cm. Without technology, use the 68–95–99.7% rule to estimate the probability that the mean height of a random sample of 1616 men is greater than 174174 cm.

Without technology (Paper 1)

Values are chosen to be a whole number of standard errors from μ\mu. Use symmetry and

P(μ±1 SE)≈0.68, P(μ±2 SE)≈0.95, P(μ±3 SE)≈0.997P(\mu \pm 1\,\mathrm{SE}) \approx 0.68,\ P(\mu \pm 2\,\mathrm{SE}) \approx 0.95,\ P(\mu \pm 3\,\mathrm{SE}) \approx 0.997

where SE=σn\mathrm{SE} = \frac{\sigma}{\sqrt{n}}.

With technology (Paper 2)

Use the normal cdf with mean μ\mu and standard deviation σn\frac{\sigma}{\sqrt{n}} (not σ\sigma). For a quantile use the inverse normal with the same inputs. Write the command and inputs in your working.

Battery lifetimes are normal with mean 480480 hours and standard deviation 3636 hours. Find the probability that the mean lifetime of a random sample of 1616 batteries is less than 470470 hours. Give your answer correct to 4 decimal places.

Round to 4 decimal places.

Quantiles work the same way. To find the value cc that a sample mean exceeds only 5% of the time, solve P(Xˉ>c)=0.05P(\bar{X} > c) = 0.05 with the inverse normal on the distribution of Xˉ\bar{X}. For the batteries, c=invNorm(0.95,480,9)≈494.80c = \text{invNorm}(0.95, 480, 9) \approx 494.80 hours.

Tip

Totals are sample means in disguise

The total of nn values is T=X1+⋯+Xn=nXˉT = X_1 + \cdots + X_n = n\bar{X}. So

P(T>t)=P(Xˉ>tn)P(T > t) = P\left(\bar{X} > \frac{t}{n}\right)

Divide the total by nn, then work with Xˉ\bar{X}.

Worked example

Will the lift be overloaded?

  • Complex unfamiliar
  • Technology active

A building's lift has a maximum safe load of 12801280 kg. The masses of adults who use it are normally distributed with mean 7676 kg and standard deviation 1414 kg.

(a) Sixteen adults get into the lift. Find the probability that their total mass exceeds the maximum safe load.

(b) The owner wants a sign "Maximum kk persons" so that, with kk adults on board, the probability of exceeding 12801280 kg is less than 1%. Find the largest such kk.

(c) State an assumption your model makes, and explain when it might not hold.

6 steps

Try it yourself first, then check the working one step at a time.

Method

  1. Rewrite a condition on a total TT as one on Xˉ=Tn\bar{X} = \frac{T}{n}.

  2. Use Xˉ∼N(μ,σ2n)\bar{X} \sim N\left(\mu, \frac{\sigma^2}{n}\right), with nn changing as the number of people changes.

  3. Test neighbouring values to find the largest one that meets the condition.

  4. Question the randomness and independence assumptions in context.

Watch out for

  • Using σ=14\sigma = 14 for the mean of 16 people.

  • Keeping SD(Xˉ)=3.5\mathrm{SD}(\bar{X}) = 3.5 when nn changes to 15.

Bags of flour have masses with mean 10051005 g and standard deviation 1212 g. For random samples of 3636 bags, find the value cc such that P(Xˉ<c)=0.01P(\bar{X} < c) = 0.01. Give your answer correct to 1 decimal place.

Round to 1 decimal place.

Exam tip

Name the distribution and the standard deviation you used

Write "Xˉ∼N(480,92)\bar{X} \sim N(480, 9^2)" before any calculator work. If you enter σ=36\sigma = 36 by mistake, the marker can still see a correct model; if you write nothing, a wrong number earns nothing.

Summary

  • For a sample mean, use Xˉ∼N(μ,σ2n)\bar{X} \sim N\left(\mu, \frac{\sigma^2}{n}\right), exact for normal XX and approximate for large nn.

  • With technology, enter the standard deviation σn\frac{\sigma}{\sqrt{n}}; without it, count standard errors and use 68–95–99.7.

  • Probabilities far from μ are much smaller for a sample mean than for one value.

  • A total of nn values is nXˉn\bar{X}: divide the limit by nn.

  • Quantiles for Xˉ\bar{X} come from the inverse normal with the same mean and standard deviation.

Finished the lesson?

Try the 3 quick checks above and the lesson completes itself, or mark it complete now.

Lesson 4 of 80% read

Save your progress with a free account