4.

3 – INTERVAL ESTIMATES FOR PROPORTIONS (Year 12 Methods)

This line
When you
HAVE 𝒑

WHEN WE HAVE THE VALUE OF 𝒑

• Population is everything / every person tested so the probability of success is 𝑝 which is the population proportion of
success based on the entire population. For example, the population proportion of PMS that are right-handed is 0.8. This
would have been a difficult task to obtain this number as everyone had to be surveyed.
• Sample is when you don’t test the entire population, just a smaller sample of 𝑛, 𝑝̂ is the sample proportion of your
𝑋
sample that had “success”, 𝑝̂ = since there were 𝑋 successes out of 𝑛 trials. This is easier than asking everyone.
𝑛
• THE PROCESS OF CALCULATING 𝒑 ̂
Whenever Sampling we need to avoid bias. The best way to avoid bias is to allocate the members of the population a
number and randomly choose the numbers of those to survey. If we ask people at a certain time (time bias), place
(location bias) our data is likely to be not representative of the entire population. Be specific when talking about bias.
1. Ask one person a Yes or No answer. Such as are you right-handed. This is a BERNOULLI trial since it is a single
event. 𝐸(𝑌) = ∑ 𝑥𝑃(𝑥) = 0 × (1 − 𝑝) + 1 × 𝑝 = 𝑝
𝑥 0 1
𝑃(𝑌 = 𝑥) 1 − 𝑝 𝑝 𝑉𝑎𝑟(𝑌) = ∑(𝑥 − 𝜇)2 𝑃(𝑥) = (0 − 𝑝)2 × (1 − 𝑝) + (1 − 𝑝)2 × 𝑝 = 𝑝(1 − 𝑝)(𝑝 + 1 − 𝑝) = 𝑝(1 − 𝑝)

2. Now ask multiple people (sample size of 𝑛) the same question, this is now BINOMIAL.
𝐸(𝑋) = 𝐸(𝑌 + 𝑌 + ⋯ + 𝑌) = 𝐸(𝑌) + 𝐸(𝑌) + ⋯ + 𝐸(𝑌) = 𝑝 + 𝑝 + ⋯ + 𝑝 = 𝑛𝑝
𝑉𝑎𝑟(𝑋) = 𝑉𝑎𝑟(𝑌 + 𝑌 + ⋯ + 𝑌) = 12 𝑉𝑎𝑟(𝑌) + 12 𝑉𝑎𝑟(𝑌) + ⋯ + 12 𝑉𝑎𝑟(𝑌) = 𝑝(1 − 𝑝) + 𝑝(1 − 𝑝) + ⋯ + 𝑝(1 − 𝑝) = 𝑛𝑝(1 − 𝑝)

3. Approximate the Binomial Distribution to a NORMAL Distribution.

The CENTRAL LIMIT THEOREM states that if you have a population with mean 𝜇 and standard deviation 𝜎 and take
sufficiently large random samples from the population with replacement, then the distribution will be
approximately normally distributed. This will hold true regardless of whether the source population is normal or
skewed, provided the sample size is sufficiently large (usually 𝑛 ≥ 30). In fact, this also holds true even if the
population is binomial, provided that 𝑛𝑝 > 10 𝑎𝑛𝑑 𝑛(1 − 𝑝) > 10, where 𝑛 is the sample size and 𝑝 is the
probability of success in the population. Therefore, the Binomial becomes Normal with same mean and variance.
𝑋
4. Change from total of yes results (𝑋) to the proportion of yes results (𝑝̂ ), by dividing by the sample size (𝑛), 𝑝̂ =
𝑛
𝑋 1 𝑛𝑝
𝐸(𝑝̂ ) = 𝐸 ( ) = 𝐸(𝑋) = =𝑝
𝑛 𝑛 𝑛

𝑋 1 2 𝑛𝑝(1 − 𝑝) 𝑝(1 − 𝑝)
𝑉𝑎𝑟(𝑝̂ ) = 𝑉𝑎𝑟 ( ) = ( ) 𝑉𝑎𝑟(𝑋) = =
𝑛 𝑛 𝑛2 𝑛
2
𝑝(1−𝑝)
So, we now have 𝑝̂ ~ 𝑁 (𝑝, ) or 𝑝̂ ~ 𝑁 (𝑝, (√𝑝(1−𝑝)) )
𝑛 𝑛

In Australia it has been found that 82% of people are right-handed. Samples of people are surveyed to ascertain the
proportion that are right-handed. Let 𝑝̂ denote the proportion of people in the sample who are right-handed.
a) State the approximate distribution of 𝑝̂ for sample sizes of 100.

b) Determine the approximate probability that in a sample of 300 people that the proportion of right handed people is
greater than 0.85

This line
When you
̂
HAVE 𝒑
̂
WHEN WE DON’T HAVE VALUE OF 𝐩, JUST A VALUE OF 𝐩
• Often we cannot test an entire population due to time and/or logistics. So, we take a sample of 𝑛 and
get the sample proportion of success 𝑝̂
• Ideally the larger the value of 𝑛, the more realistic this value of 𝑝̂ , and most likely close to the actual
value of 𝑝.
• We use 𝑝̂ as an educated first prediction for 𝑝, and with a normal approximation we can have a
p̂
confidence interval for where the actual value of 𝑝 most likely sits within a certain percentage range.
45
CONFIDENCE INTERVALS
• 𝑧 is the standardised score represented by certain Confidence Intervals, for example it is known that 95% Confidence
Interval is 𝑧 = ±1.960, since the middle 95% of scores are expected to lie within 1.960 either side of the mean. Other
common ones are 90% which is 𝑧 = ±1.645 and 99% which is 𝑧 = ±2.576
• 𝐸 is how far either side of our sample proportion (𝑝̂ ) we expect the population proportion (𝑝) to be.
• Confidence Interval is a range of values we are confident that 𝑝 lies in for a certain confidence percentage.
In a sample of 1 325 university students, 64% said that they never look at their phone while driving. Construct the 95%
confidence interval for the proportion of university students who never look at their phone while driving.
So 𝑝̂ = 0.64, 𝑛 = 1325 and we are trying to predict where 𝑝 could be.

The first 2 screens show that a 95% Confidence Interval is 𝑧 = ±1.960 (since we have a 𝜇 = 0 and 𝜎 = 1)
𝑝̂(1−𝑝̂) 0.64(1−0.64)
The last 2 screens show the 95% Confidence interval when 𝑝̂ = 0.64 and 𝜎 = √ =√
𝑛 1325

𝑝̂ (1 − 𝑝̂ ) 𝑝̂ (1 − 𝑝̂ )
𝑝̂ − 𝑧√ ≤ 𝑝 ≤ 𝑝̂ + 𝑧√
𝑛 𝑛
This is the formula to calculate this without the use of the Statistics Function on the Classpad.
0.64(1−0.64) 0.64(1−0.64)
0.64 − 1.960√ ≤ 𝑝 ≤ 0.64 + 1.960√ 0.6142 ≤ 𝑝 ≤ 0.6658
1325 1325
Since this a 95% confidence interval, if we performed the experiment 100 times, we would expect that the real population
proportion (𝑝) to be between 0.6142 and 0.6658 in 95 out of the 100 trials, and 5 times it would not (not a guarantee).
THE 95% IS NOT A PROBABLITY OR PERCENTAGE CHANCE. Ideally, you would perform several separate trials, work out the
confidence interval for each and take any overlapping sections to get an even more realistic range of values for 𝑝.

COMPARING THE POPULATION PROPORTION 𝒑 TO THE SAMPLE PROPORTION 𝒑 ̂ OBTAINED BY EXPERIMENT

This has not been consistent, so you should check the most recent WACE Exams (included in notes).
When this topic was introduced, the accepted answer was the one that was Mathematically correct, that is that based off
only one sample we cannot be certain of any conclusions as not all confidence intervals contain the true value of the
population proportion. However, it SEEMS that we are now basing our comment on whether the population proportion lies
within this confidence interval only. However, read carefully to see words like based on and refer to most recent exams.
• If the value of 𝒑 lies within the Confidence Interval obtained from the experiment 𝒑 ̂
The 95% confidence interval contains the claimed population proportion of 𝑝 = 0.8. Hence there is not
enough evidence to conclude that the proportion of rods was lower than that which was claimed.
• If the value of 𝒑 DOES NOT lie within the Confidence Interval obtained from the experiment 𝒑 ̂
The proportion of underweight noodle servings suggested by the machine specifications (𝑝 = 0.0228) is not
within the confidence interval as it is below the confidence interval. Hence the sample provides sufficient
evidence to conclude that the machine specifications is not correct at the above confidence interval.
MARGIN OF ERROR (Half the width of the Confidence Interval)
𝑝̂(1−𝑝̂)
How far from 𝑝̂ we expect the population proportion (𝑝) to be for a certain confidence is 𝑧 × 𝜎 𝐸 = 𝑧√
𝑛

A survey is completed to see who had used the city library at least once during the previous year. The researchers were given
access to data from a random sample of 159 citizens collected a few years earlier. Of these, 59 had used the city library at
least once during the previous year. Determine the margin of error for a 90% confidence interval for 𝑝 based on this sample.

An opinion poll found that 206 out of 358 people supported a policy to increase the minimum wage, from which a 99%
approximate confidence interval for the population proportion was calculated to be: (0.508, 0.643). Show how this interval
was calculated.

46
PAST WACE QUESTIONS
WACE QUESTION ONE – Calculator Assumed (2022 – Q12)

WACE QUESTION TWO – Calculator Assumed (2023 – Q7)

47
WACE QUESTION THREE – Calculator Assumed (2023 – Q12)

48