JC2 H2 Maths Bukit Timah | CLT Exact vs Approximate

Did you know? If individual observations are normally distributed, their sample mean is exactly normal for any positive sample size, even n=2. But if the population is not normal, the Central Limit Theorem usually provides an approximation for the sample mean when the sample is sufficiently large. Treating those two statements as identical can lead to unjustified answers, especially when a question provides an unusually small sample or a highly skewed population.

This JC2 H2 Mathematics Tutor Bukit Timah guide explains one specific A-Level H2 Mathematics 9758 question: when is the sampling distribution of the mean exactly normal, and when is it only approximately normal under the Central Limit Theorem? The key distinction is stated immediately, followed by two complete numerical examples, a sample-size comparison, assumptions and 36 original practice cards that separate exact calculations from approximations and unsupported conclusions.

For learners and parents in Bukit Timah, this is a foundation for interpreting normal probabilities, sample means and hypothesis-testing questions. The 2027 SEAB H2 Mathematics 9758 syllabus covers the sample mean, the exact distribution from a normal population, and use of the Central Limit Theorem when the sample size is sufficiently large (with n≥30 given as an example). It excludes normal approximation to the binomial distribution. This lesson deliberately avoids using the CLT to turn a binomial count into an assessed normal approximation.

Quick answer: first ask what distribution the population has

For independent and identically distributed observations X₁,…,Xₙ with mean μ and finite variance σ², the sample mean is X̄=(X₁+···+Xₙ)/n. Its expectation is E(X̄)=μ and its variance is Var(X̄)=σ²/n, so its standard deviation, called the standard error, is σ/√n. These moment formulas do not require a normal population, but the distribution shape is a separate question.

If each Xi is normal and the observations are independent: X̄ is exactly normal for every positive integer n.
If each Xi is non-normal, identically distributed, independent and has finite variance: the CLT says the standardised sample mean approaches a normal distribution as n grows; for sufficiently large n, normal probabilities can be approximated.
If n is small and the population is non-normal: do not assume the sample mean has a normal distribution merely from knowing μ and σ.
If the data are not independent or variance is not finite: the standard elementary CLT assumptions may fail; examine the model instead of applying a memorised n rule.
For probability calculations: standardise X̄ using σ/√n, not σ, provided a normal or justified approximately normal model is being used.
For 2027 H2 Maths: avoid substituting a normal approximation for a binomial count; that technique is expressly excluded from syllabus 9758.

Exactly normal: a small sample can still work

Suppose the individual observations are independent and each follows N(70,12²). Let the sample size be n=4. Because the population is normal, their average is exactly normally distributed with mean 70 and variance 12²/4=36, so X̄~N(70,6²). A sample size of four is not too small for this exact normality statement. The reason is the normal family itself, not a CLT approximation.

What is P(X̄>82)? Standardise with the sample-mean standard deviation 6: z=(82−70)/6=2. Therefore the exact normal-model probability is numerically P(Z>2)≈0.02275. The numeric decimal is rounded, but the distributional model for X̄ is exact under the stated independent normal assumptions. If you accidentally use σ=12, you would obtain z=1 and an incorrect probability about 0.15866.

Change only the sample size to n=16, keeping the population N(70,12²). Now X̄ is exactly N(70,3²). For the same cutoff 82, the standard score becomes 4 and the upper-tail probability becomes extremely small. The population itself did not change; averaging more independent observations made unusually high sample means much less common.

Approximately normal: a large non-normal population sample

Now imagine a different fictional population with a non-normal, mildly right-skewed distribution, mean μ=70, standard deviation σ=12 and finite variance. Suppose independent observations are sampled identically, with n=100. We do not have an exact normal distribution for X̄ from these facts. However, the Central Limit Theorem motivates the approximation X̄ approximately N(70,1.2²), because σ/√100=1.2. The approximation may be good for some underlying distributions and less accurate for others.

Using that approximation, the chance of a sample mean exceeding 72.4 is approximately P[Z>(72.4−70)/1.2]=P(Z>2)≈0.02275. The two instances of z=2 in our exact and approximate examples happen to give the same normal-tail number. But their justification differs fundamentally: in the first, exact normality follows from the population model; in the second, the probability is a CLT approximation. The answer should be labelled accordingly.

Do not read “n≥30” as a universal accuracy guarantee. It is a useful classroom guideline in the syllabus, not a theorem saying every skewed or heavy-tailed population becomes well approximated at exactly 30 observations. Accuracy depends on distribution shape, tail behaviour and assumptions. The phrase “approximately normal for sufficiently large n” communicates that limitation more faithfully than “normal whenever n reaches 30”.

Why sample means fluctuate less than individuals

A single observation has variance σ². The sum of n independent observations has variance nσ², and dividing their sum by n divides variance by n². Thus Var(X̄)=nσ²/n²=σ²/n. Taking square roots gives standard error σ/√n. This mathematical reduction is exact for independent identically distributed observations with finite variance, whether the original population is normal or not.

If the sample size is multiplied by four, the standard error is halved. For σ=12, n=25 gives standard error 12/5=2.4; n=100 gives 12/10=1.2. The change does not mean individual observations become less variable. It means the averages of repeated independent samples become more tightly concentrated around the population mean. The earlier Why Standard Error Falls by √n guide focuses on that scaling law; this page focuses on the shape and validity of the normal model.

The sample size n is not the entire justification

For independent normal observations, a sample mean of size n=2 is exactly normal. For a non-normal population, a sample mean of size n=2 need not be normal, and typically is not. Increasing n often improves a suitable normal approximation under standard finite-variance conditions. But a strongly skewed population may need more observations before a tail probability is approximated closely enough for a particular task. The location of the cutoff in the tails also matters: approximating a rare-event probability can be harder than approximating a central probability.

These points explain why H2 sampling questions normally tell you whether the population is normal or specify a sample size and conditions that justify the CLT. A learner should use the premises provided. If a question says “independent random sample from a normal population”, write “exact”. If it says “large independent sample from a non-normal population with finite variance”, write “approximately normal by CLT” and maintain the qualification in the final answer.

Mean and variance are exact even when normality is not

Suppose a non-normal population has mean 50 and standard deviation 10. For n=4 independent identically distributed observations, E(X̄)=50 and SD(X̄)=5. Those statements are valid under the moment assumptions, even though the shape of the distribution of X̄ is not supplied. Knowing the mean and standard deviation of X̄ does not automatically tell you P(X̄>60). Different non-normal populations can share those two moments yet produce different probabilities for small-sample averages.

This is a critical distinction between having enough information to calculate summary parameters and having enough information to calculate a tail probability. A calculator may ask you to enter a mean and standard deviation for a normal CDF, but it cannot certify that the random variable is normal. The student must justify that modelling step from the problem, not from the fact that two numeric parameters are available.

Why independence matters

For independent identically distributed Xi, Var(X̄)=σ²/n. If measurements move together because they share a common influence, averaging may not reduce uncertainty in the same way. For example, if every “observation” is simply the same underlying random variable X repeated n times, their average is X, and its variance is σ² rather than σ²/n. This is a counterexample to treating repeated labels as independent observations.

The elementary CLT commonly taught in H2 Mathematics assumes conditions such as independence and a common distribution with finite variance. Do not invent these assumptions merely because the word “sample” appears. A problem based on sampling without replacement from a small finite population, for instance, can involve dependence that needs a different treatment. The exact task and syllabus context determine which model is appropriate.

Why a finite variance matters

The standard form of the Central Limit Theorem taught through mean μ and standard deviation σ assumes the relevant population variance exists and is finite. If those conditions fail, a student cannot freely use σ/√n and assert a normal limit. This is not a requirement to study advanced non-normal limiting laws; it is a reminder to check that the parameters and assumptions in the stated H2 model are meaningful.

Some practical distributions are extremely heavy-tailed, and their sample means can behave differently from the familiar elementary examples. The appropriate response in a secondary education guide is modest: avoid unqualified normal probabilities if the assumptions are absent. Under the supplied normal or suitable large-sample conditions, use the standard technique confidently and label its status correctly.

Averages and totals are different random quantities

For independent Xi with mean μ and variance σ², the sum T=X₁+···+Xₙ has mean nμ and variance nσ². The average X̄=T/n has mean μ and variance σ²/n. A statement asking for P(T>c) cannot be answered by substituting c directly into a sample-mean model without dividing the threshold by n. The random variable definition must appear at the beginning of the solution.

Under independent normal observations, both the sum and the mean are exactly normal; for a suitable non-normal population and large n, each may have a CLT-motivated normal approximation after appropriate scaling. The correct means and standard deviations differ. This is why interpreting “total across 100 items” versus “average per item” matters before any graphing-calculator command.

A comparison chart for quick decisions

Stated population and sampling	What is known for X̄?	Appropriate conclusion
Independent normal Xi, n=4	Mean μ, variance σ²/4	Exactly normal
Independent normal Xi, n=100	Mean μ, variance σ²/100	Exactly normal
Independent mildly skewed Xi with finite variance, n=100	Mean μ, variance σ²/100	CLT may justify a normal approximation
Independent non-normal Xi, n=4	Mean μ, variance σ²/4	Normality not assured
Xi identical to each other (not independent)	Standard variance reduction does not follow	Do not apply the elementary independent CLT
Known mean and variance, but no distribution or sample size	Moments alone cannot fix a tail probability	More model information needed

The one-minute JC2 diagnostic

Give the learner three cases: (A) four independent normal readings; (B) one hundred independent non-normal readings from a finite-variance population; and (C) four independent non-normal readings. Ask which statements are justified. A: the mean is exactly normal. B: the mean may be approximately normal by CLT. C: normality cannot be assumed merely from those facts. Then ask for the standard error when σ=12 in each: 6, 1.2, and 6 respectively. The shape and spread questions have different answers for different reasons.

A learner who insists n must reach 30 even for a normal population has confused the exact distribution result with the CLT. One who assumes a small non-normal sample is normal because they know its mean and variance has confused moments with distribution shape. One who uses σ rather than σ/√n is confusing an individual with a sample mean. Each error has a distinct mathematical repair.

Thirty-six original exact, approximate and unsupported cases

The 36 clinics are grouped by status: first exact normal results for samples from a normal population; then CLT approximations for appropriately large samples from fictional finite-variance non-normal populations; finally small non-normal samples where no normal tail probability is justified merely by the moments. Probabilities use familiar standard-normal tail values. The worked numbers describe stipulated mathematics models, not recorded pupils, real groups or verified forecasts.

1. Exactly normal: n=1 observations, cutoff 32

Stated mathematical model. Independent observations Xi follow the same normal population N(40,4²). A sample of n=1 produces its average X̄. The question asks for the probability that X̄ exceeds 32. Before calculating, identify whether this is an exact normal result or a CLT approximation. Then state E(X̄) and the variance of X̄, not just a calculator probability.

Complete calculation. Since the population itself is normal and the observations are independent, X̄ is exactly normal with mean 40 and standard deviation 4. The corresponding variance is σ²/n=4²/1=16. Standardise the threshold: z=(32−40)/4≈-2. Therefore P(X̄>32)≈0.97725, a rounded evaluation of an exact normal-model probability.

Independent reasoning. The adjective “exact” describes the distributional normality; the displayed decimal probability is rounded. If the source population’s normality were removed, the same E(X̄)=40 and variance 16 could still hold under independent identical finite-variance sampling, but exact normality of X̄ would no longer follow. For this exercise, that exactness is justified by the stated normal parent population, even if n is small.

2. Exactly normal: n=4 observations, cutoff 47

Stated mathematical model. Independent observations Xi follow the same normal population N(47,8²). A sample of n=4 produces its average X̄. The question asks for the probability that X̄ exceeds 47. Before calculating, identify whether this is an exact normal result or a CLT approximation. Then state E(X̄) and the variance of X̄, not just a calculator probability.

Complete calculation. Since the population itself is normal and the observations are independent, X̄ is exactly normal with mean 47 and standard deviation 4. The corresponding variance is σ²/n=8²/4=16. Standardise the threshold: z=(47−47)/4≈0. Therefore P(X̄>47)≈0.5, a rounded evaluation of an exact normal-model probability.

Independent reasoning. The sample-mean standard deviation is σ/√n, not σ/n. The square root is essential. If the source population’s normality were removed, the same E(X̄)=47 and variance 16 could still hold under independent identical finite-variance sampling, but exact normality of X̄ would no longer follow. For this exercise, that exactness is justified by the stated normal parent population, even if n is small.

3. Exactly normal: n=9 observations, cutoff 60

Stated mathematical model. Independent observations Xi follow the same normal population N(54,12²). A sample of n=9 produces its average X̄. The question asks for the probability that X̄ exceeds 60. Before calculating, identify whether this is an exact normal result or a CLT approximation. Then state E(X̄) and the variance of X̄, not just a calculator probability.

Complete calculation. Since the population itself is normal and the observations are independent, X̄ is exactly normal with mean 54 and standard deviation 4. The corresponding variance is σ²/n=12²/9=16. Standardise the threshold: z=(60−54)/4≈1.5. Therefore P(X̄>60)≈0.06681, a rounded evaluation of an exact normal-model probability.

Independent reasoning. Even n=1 preserves exact normality when the underlying Xi is normal; the sample mean is then the individual observation. If the source population’s normality were removed, the same E(X̄)=54 and variance 16 could still hold under independent identical finite-variance sampling, but exact normality of X̄ would no longer follow. For this exercise, that exactness is justified by the stated normal parent population, even if n is small.

4. Exactly normal: n=16 observations, cutoff 73

Stated mathematical model. Independent observations Xi follow the same normal population N(61,16²). A sample of n=16 produces its average X̄. The question asks for the probability that X̄ exceeds 73. Before calculating, identify whether this is an exact normal result or a CLT approximation. Then state E(X̄) and the variance of X̄, not just a calculator probability.

Complete calculation. Since the population itself is normal and the observations are independent, X̄ is exactly normal with mean 61 and standard deviation 4. The corresponding variance is σ²/n=16²/16=16. Standardise the threshold: z=(73−61)/4≈3. Therefore P(X̄>73)≈0.00135, a rounded evaluation of an exact normal-model probability.

Independent reasoning. Changing n alters the spread of the sample mean, not the source population’s standard deviation. If the source population’s normality were removed, the same E(X̄)=61 and variance 16 could still hold under independent identical finite-variance sampling, but exact normality of X̄ would no longer follow. For this exercise, that exactness is justified by the stated normal parent population, even if n is small.

5. Exactly normal: n=25 observations, cutoff 66

Stated mathematical model. Independent observations Xi follow the same normal population N(68,20²). A sample of n=25 produces its average X̄. The question asks for the probability that X̄ exceeds 66. Before calculating, identify whether this is an exact normal result or a CLT approximation. Then state E(X̄) and the variance of X̄, not just a calculator probability.

Complete calculation. Since the population itself is normal and the observations are independent, X̄ is exactly normal with mean 68 and standard deviation 4. The corresponding variance is σ²/n=20²/25=16. Standardise the threshold: z=(66−68)/4≈-0.5. Therefore P(X̄>66)≈0.69146, a rounded evaluation of an exact normal-model probability.

Independent reasoning. Always identify X̄ as the random quantity before standardising the cutoff. If the source population’s normality were removed, the same E(X̄)=68 and variance 16 could still hold under independent identical finite-variance sampling, but exact normality of X̄ would no longer follow. For this exercise, that exactness is justified by the stated normal parent population, even if n is small.

6. Exactly normal: n=36 observations, cutoff 75.66667

Stated mathematical model. Independent observations Xi follow the same normal population N(75,4²). A sample of n=36 produces its average X̄. The question asks for the probability that X̄ exceeds 75.66667. Before calculating, identify whether this is an exact normal result or a CLT approximation. Then state E(X̄) and the variance of X̄, not just a calculator probability.

Complete calculation. Since the population itself is normal and the observations are independent, X̄ is exactly normal with mean 75 and standard deviation 0.66667. The corresponding variance is σ²/n=4²/36=0.44444. Standardise the threshold: z=(75.66667−75)/0.66667≈1. Therefore P(X̄>75.66667)≈0.15866, a rounded evaluation of an exact normal-model probability.

Independent reasoning. A z-score of two means two standard errors above the population mean, not two individual-observation standard deviations. If the source population’s normality were removed, the same E(X̄)=75 and variance 0.44444 could still hold under independent identical finite-variance sampling, but exact normality of X̄ would no longer follow. For this exercise, that exactness is justified by the stated normal parent population, even if n is small.

7. Exactly normal: n=49 observations, cutoff 84.85714

Stated mathematical model. Independent observations Xi follow the same normal population N(82,8²). A sample of n=49 produces its average X̄. The question asks for the probability that X̄ exceeds 84.85714. Before calculating, identify whether this is an exact normal result or a CLT approximation. Then state E(X̄) and the variance of X̄, not just a calculator probability.

Complete calculation. Since the population itself is normal and the observations are independent, X̄ is exactly normal with mean 82 and standard deviation 1.14286. The corresponding variance is σ²/n=8²/49=1.30612. Standardise the threshold: z=(84.85714−82)/1.14286≈2.5. Therefore P(X̄>84.85714)≈0.00621, a rounded evaluation of an exact normal-model probability.

Independent reasoning. For a normally distributed parent population, independent sample sums and averages remain normal without relying on a large-n CLT. If the source population’s normality were removed, the same E(X̄)=82 and variance 1.30612 could still hold under independent identical finite-variance sampling, but exact normality of X̄ would no longer follow. For this exercise, that exactness is justified by the stated normal parent population, even if n is small.

8. Exactly normal: n=64 observations, cutoff 87.5

Stated mathematical model. Independent observations Xi follow the same normal population N(89,12²). A sample of n=64 produces its average X̄. The question asks for the probability that X̄ exceeds 87.5. Before calculating, identify whether this is an exact normal result or a CLT approximation. Then state E(X̄) and the variance of X̄, not just a calculator probability.

Complete calculation. Since the population itself is normal and the observations are independent, X̄ is exactly normal with mean 89 and standard deviation 1.5. The corresponding variance is σ²/n=12²/64=2.25. Standardise the threshold: z=(87.5−89)/1.5≈-1. Therefore P(X̄>87.5)≈0.84134, a rounded evaluation of an exact normal-model probability.

Independent reasoning. Probability notation should include the actual threshold and inequality direction to avoid a wrong-tail calculation. If the source population’s normality were removed, the same E(X̄)=89 and variance 2.25 could still hold under independent identical finite-variance sampling, but exact normality of X̄ would no longer follow. For this exercise, that exactness is justified by the stated normal parent population, even if n is small.

9. Exactly normal: n=81 observations, cutoff 40.88889

Stated mathematical model. Independent observations Xi follow the same normal population N(40,16²). A sample of n=81 produces its average X̄. The question asks for the probability that X̄ exceeds 40.88889. Before calculating, identify whether this is an exact normal result or a CLT approximation. Then state E(X̄) and the variance of X̄, not just a calculator probability.

Complete calculation. Since the population itself is normal and the observations are independent, X̄ is exactly normal with mean 40 and standard deviation 1.77778. The corresponding variance is σ²/n=16²/81=3.16049. Standardise the threshold: z=(40.88889−40)/1.77778≈0.5. Therefore P(X̄>40.88889)≈0.30854, a rounded evaluation of an exact normal-model probability.

Independent reasoning. The adjective “exact” describes the distributional normality; the displayed decimal probability is rounded. If the source population’s normality were removed, the same E(X̄)=40 and variance 3.16049 could still hold under independent identical finite-variance sampling, but exact normality of X̄ would no longer follow. For this exercise, that exactness is justified by the stated normal parent population, even if n is small.

10. Exactly normal: n=100 observations, cutoff 51

Stated mathematical model. Independent observations Xi follow the same normal population N(47,20²). A sample of n=100 produces its average X̄. The question asks for the probability that X̄ exceeds 51. Before calculating, identify whether this is an exact normal result or a CLT approximation. Then state E(X̄) and the variance of X̄, not just a calculator probability.

Complete calculation. Since the population itself is normal and the observations are independent, X̄ is exactly normal with mean 47 and standard deviation 2. The corresponding variance is σ²/n=20²/100=4. Standardise the threshold: z=(51−47)/2≈2. Therefore P(X̄>51)≈0.02275, a rounded evaluation of an exact normal-model probability.

Independent reasoning. The sample-mean standard deviation is σ/√n, not σ/n. The square root is essential. If the source population’s normality were removed, the same E(X̄)=47 and variance 4 could still hold under independent identical finite-variance sampling, but exact normality of X̄ would no longer follow. For this exercise, that exactness is justified by the stated normal parent population, even if n is small.

11. Exactly normal: n=121 observations, cutoff 53.27273

Stated mathematical model. Independent observations Xi follow the same normal population N(54,4²). A sample of n=121 produces its average X̄. The question asks for the probability that X̄ exceeds 53.27273. Before calculating, identify whether this is an exact normal result or a CLT approximation. Then state E(X̄) and the variance of X̄, not just a calculator probability.

Complete calculation. Since the population itself is normal and the observations are independent, X̄ is exactly normal with mean 54 and standard deviation 0.36364. The corresponding variance is σ²/n=4²/121=0.13223. Standardise the threshold: z=(53.27273−54)/0.36364≈-2. Therefore P(X̄>53.27273)≈0.97725, a rounded evaluation of an exact normal-model probability.

Independent reasoning. Even n=1 preserves exact normality when the underlying Xi is normal; the sample mean is then the individual observation. If the source population’s normality were removed, the same E(X̄)=54 and variance 0.13223 could still hold under independent identical finite-variance sampling, but exact normality of X̄ would no longer follow. For this exercise, that exactness is justified by the stated normal parent population, even if n is small.

12. Exactly normal: n=144 observations, cutoff 61

Stated mathematical model. Independent observations Xi follow the same normal population N(61,8²). A sample of n=144 produces its average X̄. The question asks for the probability that X̄ exceeds 61. Before calculating, identify whether this is an exact normal result or a CLT approximation. Then state E(X̄) and the variance of X̄, not just a calculator probability.

Complete calculation. Since the population itself is normal and the observations are independent, X̄ is exactly normal with mean 61 and standard deviation 0.66667. The corresponding variance is σ²/n=8²/144=0.44444. Standardise the threshold: z=(61−61)/0.66667≈0. Therefore P(X̄>61)≈0.5, a rounded evaluation of an exact normal-model probability.

Independent reasoning. Changing n alters the spread of the sample mean, not the source population’s standard deviation. If the source population’s normality were removed, the same E(X̄)=61 and variance 0.44444 could still hold under independent identical finite-variance sampling, but exact normality of X̄ would no longer follow. For this exercise, that exactness is justified by the stated normal parent population, even if n is small.

13. Exactly normal: n=169 observations, cutoff 69.38462

Stated mathematical model. Independent observations Xi follow the same normal population N(68,12²). A sample of n=169 produces its average X̄. The question asks for the probability that X̄ exceeds 69.38462. Before calculating, identify whether this is an exact normal result or a CLT approximation. Then state E(X̄) and the variance of X̄, not just a calculator probability.

Complete calculation. Since the population itself is normal and the observations are independent, X̄ is exactly normal with mean 68 and standard deviation 0.92308. The corresponding variance is σ²/n=12²/169=0.85207. Standardise the threshold: z=(69.38462−68)/0.92308≈1.5. Therefore P(X̄>69.38462)≈0.06681, a rounded evaluation of an exact normal-model probability.

Independent reasoning. Always identify X̄ as the random quantity before standardising the cutoff. If the source population’s normality were removed, the same E(X̄)=68 and variance 0.85207 could still hold under independent identical finite-variance sampling, but exact normality of X̄ would no longer follow. For this exercise, that exactness is justified by the stated normal parent population, even if n is small.

14. Exactly normal: n=196 observations, cutoff 78.42857

Stated mathematical model. Independent observations Xi follow the same normal population N(75,16²). A sample of n=196 produces its average X̄. The question asks for the probability that X̄ exceeds 78.42857. Before calculating, identify whether this is an exact normal result or a CLT approximation. Then state E(X̄) and the variance of X̄, not just a calculator probability.

Complete calculation. Since the population itself is normal and the observations are independent, X̄ is exactly normal with mean 75 and standard deviation 1.14286. The corresponding variance is σ²/n=16²/196=1.30612. Standardise the threshold: z=(78.42857−75)/1.14286≈3. Therefore P(X̄>78.42857)≈0.00135, a rounded evaluation of an exact normal-model probability.

Independent reasoning. A z-score of two means two standard errors above the population mean, not two individual-observation standard deviations. If the source population’s normality were removed, the same E(X̄)=75 and variance 1.30612 could still hold under independent identical finite-variance sampling, but exact normality of X̄ would no longer follow. For this exercise, that exactness is justified by the stated normal parent population, even if n is small.

15. Exactly normal: n=225 observations, cutoff 81.33333

Stated mathematical model. Independent observations Xi follow the same normal population N(82,20²). A sample of n=225 produces its average X̄. The question asks for the probability that X̄ exceeds 81.33333. Before calculating, identify whether this is an exact normal result or a CLT approximation. Then state E(X̄) and the variance of X̄, not just a calculator probability.

Complete calculation. Since the population itself is normal and the observations are independent, X̄ is exactly normal with mean 82 and standard deviation 1.33333. The corresponding variance is σ²/n=20²/225=1.77778. Standardise the threshold: z=(81.33333−82)/1.33333≈-0.5. Therefore P(X̄>81.33333)≈0.69146, a rounded evaluation of an exact normal-model probability.

Independent reasoning. For a normally distributed parent population, independent sample sums and averages remain normal without relying on a large-n CLT. If the source population’s normality were removed, the same E(X̄)=82 and variance 1.77778 could still hold under independent identical finite-variance sampling, but exact normality of X̄ would no longer follow. For this exercise, that exactness is justified by the stated normal parent population, even if n is small.

16. Exactly normal: n=256 observations, cutoff 89.25

Stated mathematical model. Independent observations Xi follow the same normal population N(89,4²). A sample of n=256 produces its average X̄. The question asks for the probability that X̄ exceeds 89.25. Before calculating, identify whether this is an exact normal result or a CLT approximation. Then state E(X̄) and the variance of X̄, not just a calculator probability.

Complete calculation. Since the population itself is normal and the observations are independent, X̄ is exactly normal with mean 89 and standard deviation 0.25. The corresponding variance is σ²/n=4²/256=0.0625. Standardise the threshold: z=(89.25−89)/0.25≈1. Therefore P(X̄>89.25)≈0.15866, a rounded evaluation of an exact normal-model probability.

Independent reasoning. Probability notation should include the actual threshold and inequality direction to avoid a wrong-tail calculation. If the source population’s normality were removed, the same E(X̄)=89 and variance 0.0625 could still hold under independent identical finite-variance sampling, but exact normality of X̄ would no longer follow. For this exercise, that exactness is justified by the stated normal parent population, even if n is small.

17. CLT approximation: n=64 non-normal observations

Stated hypothetical sampling model. Xi are independent, identically distributed observations from a non-normal, moderately skewed population with finite mean μ=30 and standard deviation σ=8. A random sample of n=64 is taken and averaged. The question asks for P(X̄>31). Explain why a normal distribution for X̄ is not exact from the given population description and what approximation may be used for a sufficiently large n.

Calculation with its qualification. The mean of X̄ is exactly 30 and its standard error is exactly 8/√64≈1 under the independent sampling model. For sufficiently large n, the CLT motivates the approximation X̄≈N(30, 1²). The threshold standardises to z≈1, so the approximate normal tail is P(X̄>31)≈0.15866. The displayed number is a CLT approximation; the exact finite-n tail is not determined by μ and σ alone.

Check the epistemic status. The approximation concerns averages of repeated samples, not the distribution of individual observations. If the original population were known to be normal instead, this same form of distribution for the sample mean would be exact. Here the parent population is expressly non-normal, and the approximation’s quality depends on that population’s actual shape. A tutor should ask the student to retain the word “approximately” in the final sentence rather than silently promoting the model into a proven identity.

18. CLT approximation: n=81 non-normal observations

Stated hypothetical sampling model. Xi are independent, identically distributed observations from a non-normal, moderately skewed population with finite mean μ=40 and standard deviation σ=12. A random sample of n=81 is taken and averaged. The question asks for P(X̄>43.33333). Explain why a normal distribution for X̄ is not exact from the given population description and what approximation may be used for a sufficiently large n.

Calculation with its qualification. The mean of X̄ is exactly 40 and its standard error is exactly 12/√81≈1.33333 under the independent sampling model. For sufficiently large n, the CLT motivates the approximation X̄≈N(40, 1.33333²). The threshold standardises to z≈2.5, so the approximate normal tail is P(X̄>43.33333)≈0.00621. The displayed number is a CLT approximation; the exact finite-n tail is not determined by μ and σ alone.

Check the epistemic status. A large sample can make a normal approximation reasonable, but its quality is not guaranteed by an n threshold alone. If the original population were known to be normal instead, this same form of distribution for the sample mean would be exact. Here the parent population is expressly non-normal, and the approximation’s quality depends on that population’s actual shape. A tutor should ask the student to retain the word “approximately” in the final sentence rather than silently promoting the model into a proven identity.

19. CLT approximation: n=100 non-normal observations

Stated hypothetical sampling model. Xi are independent, identically distributed observations from a non-normal, moderately skewed population with finite mean μ=50 and standard deviation σ=16. A random sample of n=100 is taken and averaged. The question asks for P(X̄>48.4). Explain why a normal distribution for X̄ is not exact from the given population description and what approximation may be used for a sufficiently large n.

Calculation with its qualification. The mean of X̄ is exactly 50 and its standard error is exactly 16/√100≈1.6 under the independent sampling model. For sufficiently large n, the CLT motivates the approximation X̄≈N(50, 1.6²). The threshold standardises to z≈-1, so the approximate normal tail is P(X̄>48.4)≈0.84134. The displayed number is a CLT approximation; the exact finite-n tail is not determined by μ and σ alone.

Check the epistemic status. A far-tail probability can be more sensitive to the population’s shape than a central probability. If the original population were known to be normal instead, this same form of distribution for the sample mean would be exact. Here the parent population is expressly non-normal, and the approximation’s quality depends on that population’s actual shape. A tutor should ask the student to retain the word “approximately” in the final sentence rather than silently promoting the model into a proven identity.

20. CLT approximation: n=121 non-normal observations

Stated hypothetical sampling model. Xi are independent, identically distributed observations from a non-normal, moderately skewed population with finite mean μ=60 and standard deviation σ=20. A random sample of n=121 is taken and averaged. The question asks for P(X̄>60.90909). Explain why a normal distribution for X̄ is not exact from the given population description and what approximation may be used for a sufficiently large n.

Calculation with its qualification. The mean of X̄ is exactly 60 and its standard error is exactly 20/√121≈1.81818 under the independent sampling model. For sufficiently large n, the CLT motivates the approximation X̄≈N(60, 1.81818²). The threshold standardises to z≈0.5, so the approximate normal tail is P(X̄>60.90909)≈0.30854. The displayed number is a CLT approximation; the exact finite-n tail is not determined by μ and σ alone.

Check the epistemic status. The exact moment formulas still apply even though the distributional shape is approximate. If the original population were known to be normal instead, this same form of distribution for the sample mean would be exact. Here the parent population is expressly non-normal, and the approximation’s quality depends on that population’s actual shape. A tutor should ask the student to retain the word “approximately” in the final sentence rather than silently promoting the model into a proven identity.

21. CLT approximation: n=144 non-normal observations

Stated hypothetical sampling model. Xi are independent, identically distributed observations from a non-normal, moderately skewed population with finite mean μ=70 and standard deviation σ=8. A random sample of n=144 is taken and averaged. The question asks for P(X̄>71.33333). Explain why a normal distribution for X̄ is not exact from the given population description and what approximation may be used for a sufficiently large n.

Calculation with its qualification. The mean of X̄ is exactly 70 and its standard error is exactly 8/√144≈0.66667 under the independent sampling model. For sufficiently large n, the CLT motivates the approximation X̄≈N(70, 0.66667²). The threshold standardises to z≈2, so the approximate normal tail is P(X̄>71.33333)≈0.02275. The displayed number is a CLT approximation; the exact finite-n tail is not determined by μ and σ alone.

Check the epistemic status. The distribution is not being asserted to become exactly normal at this particular finite n. If the original population were known to be normal instead, this same form of distribution for the sample mean would be exact. Here the parent population is expressly non-normal, and the approximation’s quality depends on that population’s actual shape. A tutor should ask the student to retain the word “approximately” in the final sentence rather than silently promoting the model into a proven identity.

22. CLT approximation: n=169 non-normal observations

Stated hypothetical sampling model. Xi are independent, identically distributed observations from a non-normal, moderately skewed population with finite mean μ=80 and standard deviation σ=12. A random sample of n=169 is taken and averaged. The question asks for P(X̄>78.15385). Explain why a normal distribution for X̄ is not exact from the given population description and what approximation may be used for a sufficiently large n.

Calculation with its qualification. The mean of X̄ is exactly 80 and its standard error is exactly 12/√169≈0.92308 under the independent sampling model. For sufficiently large n, the CLT motivates the approximation X̄≈N(80, 0.92308²). The threshold standardises to z≈-2, so the approximate normal tail is P(X̄>78.15385)≈0.97725. The displayed number is a CLT approximation; the exact finite-n tail is not determined by μ and σ alone.

Check the epistemic status. Independence, identical sampling and finite variance are essential premises for the elementary CLT use stated here. If the original population were known to be normal instead, this same form of distribution for the sample mean would be exact. Here the parent population is expressly non-normal, and the approximation’s quality depends on that population’s actual shape. A tutor should ask the student to retain the word “approximately” in the final sentence rather than silently promoting the model into a proven identity.

23. CLT approximation: n=196 non-normal observations

Stated hypothetical sampling model. Xi are independent, identically distributed observations from a non-normal, moderately skewed population with finite mean μ=90 and standard deviation σ=16. A random sample of n=196 is taken and averaged. The question asks for P(X̄>90). Explain why a normal distribution for X̄ is not exact from the given population description and what approximation may be used for a sufficiently large n.

Calculation with its qualification. The mean of X̄ is exactly 90 and its standard error is exactly 16/√196≈1.14286 under the independent sampling model. For sufficiently large n, the CLT motivates the approximation X̄≈N(90, 1.14286²). The threshold standardises to z≈0, so the approximate normal tail is P(X̄>90)≈0.5. The displayed number is a CLT approximation; the exact finite-n tail is not determined by μ and σ alone.

Check the epistemic status. A mildly skewed hypothetical population is stipulated; actual model quality would need more information or diagnostics. If the original population were known to be normal instead, this same form of distribution for the sample mean would be exact. Here the parent population is expressly non-normal, and the approximation’s quality depends on that population’s actual shape. A tutor should ask the student to retain the word “approximately” in the final sentence rather than silently promoting the model into a proven identity.

24. CLT approximation: n=225 non-normal observations

Stated hypothetical sampling model. Xi are independent, identically distributed observations from a non-normal, moderately skewed population with finite mean μ=30 and standard deviation σ=20. A random sample of n=225 is taken and averaged. The question asks for P(X̄>32). Explain why a normal distribution for X̄ is not exact from the given population description and what approximation may be used for a sufficiently large n.

Calculation with its qualification. The mean of X̄ is exactly 30 and its standard error is exactly 20/√225≈1.33333 under the independent sampling model. For sufficiently large n, the CLT motivates the approximation X̄≈N(30, 1.33333²). The threshold standardises to z≈1.5, so the approximate normal tail is P(X̄>32)≈0.06681. The displayed number is a CLT approximation; the exact finite-n tail is not determined by μ and σ alone.

Check the epistemic status. The resulting z-score is standardised using the standard error σ/√n, not the individual σ. If the original population were known to be normal instead, this same form of distribution for the sample mean would be exact. Here the parent population is expressly non-normal, and the approximation’s quality depends on that population’s actual shape. A tutor should ask the student to retain the word “approximately” in the final sentence rather than silently promoting the model into a proven identity.

25. CLT approximation: n=256 non-normal observations

Stated hypothetical sampling model. Xi are independent, identically distributed observations from a non-normal, moderately skewed population with finite mean μ=40 and standard deviation σ=8. A random sample of n=256 is taken and averaged. The question asks for P(X̄>41.5). Explain why a normal distribution for X̄ is not exact from the given population description and what approximation may be used for a sufficiently large n.

Calculation with its qualification. The mean of X̄ is exactly 40 and its standard error is exactly 8/√256≈0.5 under the independent sampling model. For sufficiently large n, the CLT motivates the approximation X̄≈N(40, 0.5²). The threshold standardises to z≈3, so the approximate normal tail is P(X̄>41.5)≈0.00135. The displayed number is a CLT approximation; the exact finite-n tail is not determined by μ and σ alone.

Check the epistemic status. The displayed tail area is from a standard-normal approximation, and the approximation qualifier must follow it into the conclusion. If the original population were known to be normal instead, this same form of distribution for the sample mean would be exact. Here the parent population is expressly non-normal, and the approximation’s quality depends on that population’s actual shape. A tutor should ask the student to retain the word “approximately” in the final sentence rather than silently promoting the model into a proven identity.

26. CLT approximation: n=289 non-normal observations

Stated hypothetical sampling model. Xi are independent, identically distributed observations from a non-normal, moderately skewed population with finite mean μ=50 and standard deviation σ=12. A random sample of n=289 is taken and averaged. The question asks for P(X̄>49.64706). Explain why a normal distribution for X̄ is not exact from the given population description and what approximation may be used for a sufficiently large n.

Calculation with its qualification. The mean of X̄ is exactly 50 and its standard error is exactly 12/√289≈0.70588 under the independent sampling model. For sufficiently large n, the CLT motivates the approximation X̄≈N(50, 0.70588²). The threshold standardises to z≈-0.5, so the approximate normal tail is P(X̄>49.64706)≈0.69146. The displayed number is a CLT approximation; the exact finite-n tail is not determined by μ and σ alone.

Check the epistemic status. Increasing n will generally concentrate X̄ more tightly about μ, but it does not change the population distribution of Xi. If the original population were known to be normal instead, this same form of distribution for the sample mean would be exact. Here the parent population is expressly non-normal, and the approximation’s quality depends on that population’s actual shape. A tutor should ask the student to retain the word “approximately” in the final sentence rather than silently promoting the model into a proven identity.

27. CLT approximation: n=324 non-normal observations

Stated hypothetical sampling model. Xi are independent, identically distributed observations from a non-normal, moderately skewed population with finite mean μ=60 and standard deviation σ=16. A random sample of n=324 is taken and averaged. The question asks for P(X̄>60.88889). Explain why a normal distribution for X̄ is not exact from the given population description and what approximation may be used for a sufficiently large n.

Calculation with its qualification. The mean of X̄ is exactly 60 and its standard error is exactly 16/√324≈0.88889 under the independent sampling model. For sufficiently large n, the CLT motivates the approximation X̄≈N(60, 0.88889²). The threshold standardises to z≈1, so the approximate normal tail is P(X̄>60.88889)≈0.15866. The displayed number is a CLT approximation; the exact finite-n tail is not determined by μ and σ alone.

Check the epistemic status. Without information about shape, it is unsafe to claim that a numerical finite-sample normal approximation has known accuracy. If the original population were known to be normal instead, this same form of distribution for the sample mean would be exact. Here the parent population is expressly non-normal, and the approximation’s quality depends on that population’s actual shape. A tutor should ask the student to retain the word “approximately” in the final sentence rather than silently promoting the model into a proven identity.

28. CLT approximation: n=400 non-normal observations

Stated hypothetical sampling model. Xi are independent, identically distributed observations from a non-normal, moderately skewed population with finite mean μ=70 and standard deviation σ=20. A random sample of n=400 is taken and averaged. The question asks for P(X̄>72.5). Explain why a normal distribution for X̄ is not exact from the given population description and what approximation may be used for a sufficiently large n.

Calculation with its qualification. The mean of X̄ is exactly 70 and its standard error is exactly 20/√400≈1 under the independent sampling model. For sufficiently large n, the CLT motivates the approximation X̄≈N(70, 1²). The threshold standardises to z≈2.5, so the approximate normal tail is P(X̄>72.5)≈0.00621. The displayed number is a CLT approximation; the exact finite-n tail is not determined by μ and σ alone.

Check the epistemic status. A model may give a convenient approximate percentage without proving a particular outcome will occur. If the original population were known to be normal instead, this same form of distribution for the sample mean would be exact. Here the parent population is expressly non-normal, and the approximation’s quality depends on that population’s actual shape. A tutor should ask the student to retain the word “approximately” in the final sentence rather than silently promoting the model into a proven identity.

29. CLT approximation: n=625 non-normal observations

Stated hypothetical sampling model. Xi are independent, identically distributed observations from a non-normal, moderately skewed population with finite mean μ=80 and standard deviation σ=8. A random sample of n=625 is taken and averaged. The question asks for P(X̄>79.68). Explain why a normal distribution for X̄ is not exact from the given population description and what approximation may be used for a sufficiently large n.

Calculation with its qualification. The mean of X̄ is exactly 80 and its standard error is exactly 8/√625≈0.32 under the independent sampling model. For sufficiently large n, the CLT motivates the approximation X̄≈N(80, 0.32²). The threshold standardises to z≈-1, so the approximate normal tail is P(X̄>79.68)≈0.84134. The displayed number is a CLT approximation; the exact finite-n tail is not determined by μ and σ alone.

Check the epistemic status. The probability addresses the sample mean; if the question asked for a total, it would need new parameters. If the original population were known to be normal instead, this same form of distribution for the sample mean would be exact. Here the parent population is expressly non-normal, and the approximation’s quality depends on that population’s actual shape. A tutor should ask the student to retain the word “approximately” in the final sentence rather than silently promoting the model into a proven identity.

30. CLT approximation: n=900 non-normal observations

Stated hypothetical sampling model. Xi are independent, identically distributed observations from a non-normal, moderately skewed population with finite mean μ=90 and standard deviation σ=12. A random sample of n=900 is taken and averaged. The question asks for P(X̄>90.2). Explain why a normal distribution for X̄ is not exact from the given population description and what approximation may be used for a sufficiently large n.

Calculation with its qualification. The mean of X̄ is exactly 90 and its standard error is exactly 12/√900≈0.4 under the independent sampling model. For sufficiently large n, the CLT motivates the approximation X̄≈N(90, 0.4²). The threshold standardises to z≈0.5, so the approximate normal tail is P(X̄>90.2)≈0.30854. The displayed number is a CLT approximation; the exact finite-n tail is not determined by μ and σ alone.

Check the epistemic status. Do not confuse this sampling-mean approximation with the normal approximation to a binomial distribution excluded from the 2027 H2 syllabus. If the original population were known to be normal instead, this same form of distribution for the sample mean would be exact. Here the parent population is expressly non-normal, and the approximation’s quality depends on that population’s actual shape. A tutor should ask the student to retain the word “approximately” in the final sentence rather than silently promoting the model into a proven identity.

31. Small non-normal sample: do not invent a normal tail

Read the assumptions. A hypothetical non-normal population has mean μ=45 and standard deviation σ=6. Independent identically distributed observations are sampled with n=2, and the question asks about P(X̄>49.24264). The population’s exact shape is not supplied. Decide what can be calculated from the given information and what cannot be inferred with confidence.

Correct result. Under independent sampling, E(X̄)=45 and SD(X̄)=σ/√n=4.24264. These are exact first and second moments. But because n=2 is small and the parent population is not normal, there is no justification from the stated facts alone for declaring X̄ normal or for reporting a numerical standard-normal tail probability. Different parent distributions with these same moments can give different small-sample tails.

What to say instead. An exam-ready response distinguishes the known mean and standard error from the unknown distribution shape. If a larger sample or an explicitly justified approximation were given, a CLT method might become appropriate. If the parent population were normal, the sample mean would be exactly normal even for this n. The purpose of this clinic is to practise refusing an unjustified calculator shortcut, not to leave the algebra incomplete.

32. Small non-normal sample: do not invent a normal tail

Read the assumptions. A hypothetical non-normal population has mean μ=50 and standard deviation σ=8. Independent identically distributed observations are sampled with n=3, and the question asks about P(X̄>54.6188). The population’s exact shape is not supplied. Decide what can be calculated from the given information and what cannot be inferred with confidence.

Correct result. Under independent sampling, E(X̄)=50 and SD(X̄)=σ/√n=4.6188. These are exact first and second moments. But because n=3 is small and the parent population is not normal, there is no justification from the stated facts alone for declaring X̄ normal or for reporting a numerical standard-normal tail probability. Different parent distributions with these same moments can give different small-sample tails.

33. Small non-normal sample: do not invent a normal tail

Read the assumptions. A hypothetical non-normal population has mean μ=55 and standard deviation σ=10. Independent identically distributed observations are sampled with n=4, and the question asks about P(X̄>60). The population’s exact shape is not supplied. Decide what can be calculated from the given information and what cannot be inferred with confidence.

Correct result. Under independent sampling, E(X̄)=55 and SD(X̄)=σ/√n=5. These are exact first and second moments. But because n=4 is small and the parent population is not normal, there is no justification from the stated facts alone for declaring X̄ normal or for reporting a numerical standard-normal tail probability. Different parent distributions with these same moments can give different small-sample tails.

34. Small non-normal sample: do not invent a normal tail

Read the assumptions. A hypothetical non-normal population has mean μ=60 and standard deviation σ=12. Independent identically distributed observations are sampled with n=5, and the question asks about P(X̄>65.36656). The population’s exact shape is not supplied. Decide what can be calculated from the given information and what cannot be inferred with confidence.

Correct result. Under independent sampling, E(X̄)=60 and SD(X̄)=σ/√n=5.36656. These are exact first and second moments. But because n=5 is small and the parent population is not normal, there is no justification from the stated facts alone for declaring X̄ normal or for reporting a numerical standard-normal tail probability. Different parent distributions with these same moments can give different small-sample tails.

35. Small non-normal sample: do not invent a normal tail

Read the assumptions. A hypothetical non-normal population has mean μ=65 and standard deviation σ=14. Independent identically distributed observations are sampled with n=6, and the question asks about P(X̄>70.71548). The population’s exact shape is not supplied. Decide what can be calculated from the given information and what cannot be inferred with confidence.

Correct result. Under independent sampling, E(X̄)=65 and SD(X̄)=σ/√n=5.71548. These are exact first and second moments. But because n=6 is small and the parent population is not normal, there is no justification from the stated facts alone for declaring X̄ normal or for reporting a numerical standard-normal tail probability. Different parent distributions with these same moments can give different small-sample tails.

36. Small non-normal sample: do not invent a normal tail

Read the assumptions. A hypothetical non-normal population has mean μ=70 and standard deviation σ=16. Independent identically distributed observations are sampled with n=8, and the question asks about P(X̄>75.65685). The population’s exact shape is not supplied. Decide what can be calculated from the given information and what cannot be inferred with confidence.

Correct result. Under independent sampling, E(X̄)=70 and SD(X̄)=σ/√n=5.65685. These are exact first and second moments. But because n=8 is small and the parent population is not normal, there is no justification from the stated facts alone for declaring X̄ normal or for reporting a numerical standard-normal tail probability. Different parent distributions with these same moments can give different small-sample tails.

Ten fast checks of exactness and assumptions

1. Independent Xi~N(70,12²), n=4: X̄ is exactly N(70,6²).
2. For that exact model, P(X̄>82)≈0.02275.
3. Same normal parent, n=16: X̄ is exactly N(70,3²).
4. Non-normal independent observations with finite variance and large n: CLT can justify an approximation, not exact normality.
5. Non-normal parent, n=4: a normal sample-mean tail is not automatically justified.
6. For independent identically distributed observations, E(X̄)=μ.
7. Under the same assumptions, Var(X̄)=σ²/n.
8. Standard error is σ/√n, not σ/n.
9. If four copies of the same random variable are perfectly dependent, the average has the same variance as that variable, not one-quarter.
10. The 2027 H2 syllabus excludes normal approximation to the binomial distribution.

An exam-ready five-day revision structure

Day one: distinguish an individual observation from its sample mean and from a sample total. Day two: calculate means and standard errors under independent sampling, with no normal probabilities yet. Day three: use small samples from a normal population and label why the result is exact. Day four: use adequately large samples from a non-normal finite-variance population and state the CLT approximation with its assumptions. Day five: mix those cases with small non-normal samples, missing independence and unknown variance, and identify when a numeric normal tail cannot be justified.

In a three-person tutorial, one student identifies the source population, another finds the distribution or moments of X̄, and the third checks whether the word “exact” or “approximately” is warranted. Rotate the roles. This shifts the focus from a calculator menu to the mathematical evidence required for the menu to be appropriate. Learners who can explain why the same z-score can come from either an exact or an approximate normal model have grasped the central distinction.

Frequently asked questions

Is the sample mean normal for every sample size?

If the independent observations come from a normal population, yes. If they come from a non-normal population, not generally; the CLT may provide an approximation as sample size increases.

Is n≥30 a guarantee of normality?

No. The 2027 syllabus gives n≥30 as an example of sufficiently large sampling for classroom CLT use, but approximation quality depends on the population and problem.

Are E(X̄) and Var(X̄) exact when Xi are non-normal?

Under independent identically distributed sampling with finite variance, yes: E(X̄)=μ and Var(X̄)=σ²/n do not require a normal parent distribution.

What is the difference between σ and σ/√n?

σ is the spread of individual observations. σ/√n is the standard deviation of repeated sample means under the stated independent sampling model.

Can small non-normal samples have approximately normal means?

Some can, depending on the population, but the sample size alone and two moments do not guarantee it. The model needs justification.

What happens without independence?

The usual variance reduction and elementary CLT assumptions may fail. Read the sampling design instead of assuming independent observations.

Does normality of Xi make X̄ exactly normal?

Yes, for independent normal observations, including n=1 and n=2. The resulting mean and variance of X̄ follow the standard formulas.

Is normal approximation of a binomial part of H2 9758?

No. The 2027 H2 Mathematics syllabus explicitly excludes normal approximation to the binomial distribution, which this guide does not use for its worked probabilities.

Links to related tuition and curriculum guides

A Hole or a Vertical Asymptote?
Quadratic Inverse: Why Restrict the Domain?
Cross Product: Triangle Area or Parallelogram?
CLT: Exact or Approximate Normal?: current guide

For the broader sampling route read Bukit Timah H2 Normal Distribution, Sampling and CLT and the narrower Why Standard Error Falls by √n. The Bukit Timah JC2 tuition hub provides location context; the immutable Clementi three-pupil Mathematics tutorial reference illustrates the teaching model at another year level.

Official source: SEAB Singapore–Cambridge H2 Mathematics 9758, 2027 syllabus, Sampling and Normal Distribution. All 36 cases are original mathematical models rather than official past-year exam questions or verified student performance. When the normal model is approximate, its numerical probabilities are also approximate and must be described as such.

JC2 H2 Mathematics Tutor Bukit Timah | Central Limit Theorem: Exact or Approximate Normal?