The Intuition of Standardising

Why does $Z = \dfrac{X-\mu}{\sigma}$ work?

VCE Methods Β· Normal Distribution Β· Sample Proportions

πŸ“ Designed for use with the TI-Nspire CX II

πŸ“Œ The Problem We Are Solving (Example 12)

Suppose 60% of people hold a driver's licence, and a random sample of size $n$ is selected. If $\Pr(\hat{P} < 0.58) = 0.3446$, find $n$.

What we already know

  • $p = 0.60$ β€” the population proportion
  • $\hat{P} \sim N(\mu,\sigma^2)$ β€” approximately normal
  • $\mu = p = 0.60$
  • $\sigma = \sqrt{\dfrac{p(1-p)}{n}} = \sqrt{\dfrac{0.24}{n}}$

⚠️ The calculator's invNorm works on the standard normal $Z$, so we must convert $\hat{P}$ into $Z$ first.

πŸ€” Where Students Often Get Stuck

"Why do we subtract $\mu$ and divide by $\sigma$? Where does this formula come from?"

Many students treat $Z = \dfrac{X-\mu}{\sigma}$ as a formula to memorise.

In fact, it is just two geometric moves:

  • Step 1 β€” Subtract $\mu$ β†’ shift the centre to 0
  • Step 2 β€” Divide by $\sigma$ β†’ change units to "number of standard deviations"
πŸ’‘ By the end of these slides, $Z$ will not look like a formula. It will be the same data described in another way.

🎯 Intuition 1 β€” Subtract $\mu$ to Shift

The original distribution $X \sim N(\mu, \sigma^2)$ has its bell centred at $\mu$.

ΞΌ x X ~ N(ΞΌ, σ²) βˆ’ΞΌ 0 X βˆ’ ΞΌ ~ N(0, σ²)
After shifting, the shape is unchanged; only the centre moved from $\mu$ to $0$. The standard deviation is still $\sigma$.

πŸ“ Intuition 2 β€” Divide by $\sigma$ to Rescale

The bell is still "wide or narrow" depending on $\sigma$. We want every normal distribution to map to one universal curve.

0 X βˆ’ ΞΌ ~ N(0, σ²) width = Οƒ Γ· Οƒ 0 Z ~ N(0, 1) width = 1
After dividing by $\sigma$, the horizontal unit changes from "$X$ units" to "numbers of standard deviations". A value of $z = 1$ means "exactly one standard deviation above the mean".

πŸ’Ž The Key Insight β€” A New "Ruler"

Imagine two students sitting different exams:

StudentScore $X$Class $\mu$Class $\sigma$$Z = (X-\mu)/\sigma$
Alice857010+1.5
Bob62508+1.5
Their raw scores are very different (85 vs 62), but their $Z$ scores match. Both students are "1.5 standard deviations above their class mean" β€” the same relative position.

πŸ“Œ Standardising lets us compare different normal distributions on a common scale.

πŸ“ The Formula and Probability Invariance

$$Z = \frac{X - \mu}{\sigma} \;\sim\; N(0,\,1)$$

Translating probability statements

Because standardising only changes the ruler, not the event:

$$\Pr(X < a) \;=\; \Pr\!\left(Z < \frac{a-\mu}{\sigma}\right)$$
The left side asks "the probability that $X$ falls below $a$".
The right side asks "the probability that $Z$ falls below $\tfrac{a-\mu}{\sigma}$".
They describe the same event in two coordinate systems.

This is precisely why a single standard normal table β€” or a single calculator function β€” solves any normal problem.

πŸ” Back to Example 12 β€” Standardising the Statement

We want $n$ such that $\Pr(\hat{P} < 0.58) = 0.3446$.

1Standardise $\hat{P}$: $$\Pr(\hat{P} < 0.58) = \Pr\!\left(Z < \frac{0.58 - 0.60}{\sqrt{0.24/n}}\right)$$
2Simplify the numerator: $$\Pr\!\left(Z < \frac{-0.02}{\sqrt{0.24/n}}\right) = 0.3446$$

The right side is now a pure $Z$ statement β€” ready for the calculator.

🧠 Strategy: the probability is on the left, the unknown $z$ is on the right. We ask "which $z$ makes the cumulative probability equal to $0.3446$?"

πŸ–© TI-Nspire CX II β€” Inverse Normal

Since $0.3446 < 0.5$, the corresponding $z$ value is negative (left of the mean).

How to call invNorm

menu β†’ 5: Probability
   β†’ 5: Distributions
   β†’ 3: Inverse Normal

Area: 0.3446
ΞΌ: 0
Οƒ: 1

β†’ z β‰ˆ βˆ’0.400

What does $-0.400$ mean?

It is the standard-normal quantile that satisfies $\Pr(Z < z) = 0.3446$.

So the two sides agree:

$$\frac{-0.02}{\sqrt{0.24/n}} = -0.400$$
πŸ’‘ Rule of thumb: any inverse-probability lookup is always done on the standard normal $N(0,1)$.

✏️ Solving the Equation for $n$

1Start from $\dfrac{-0.02}{\sqrt{0.24/n}} = -0.400$ and multiply both sides by the square root: $$-0.02 = -0.400 \cdot \sqrt{\frac{0.24}{n}}$$
2Divide both sides by $-0.400$: $$\sqrt{\frac{0.24}{n}} = 0.05$$
3Square both sides: $$\frac{0.24}{n} = 0.0025$$
4Solve for $n$: $$n = \frac{0.24}{0.0025} = \boxed{96}$$

βœ… The required sample size is 96 people.

πŸ—ΊοΈ The Whole Roadmap

1
Model
$\hat{P} \sim N(\mu,\sigma^2)$
$\mu=p,\ \sigma=\sqrt{\tfrac{p(1-p)}{n}}$
2
Standardise
Rewrite as $\Pr(Z < z) = \text{given}$
by translating $\hat{P}$ to $Z$
3
Inverse $z$
Use invNorm(area, 0, 1)
to recover the matching $z$
4
Substitute back
$z = \dfrac{a-\mu}{\sigma}$
becomes an equation in $n$
5
Solve algebraically
Rearrange and find $n$
6
Sanity check
Is $n$ large enough
for the normal approximation?
🎯 Standardising is the bridge from the question's $X$ or $\hat{P}$ to the calculator's $Z$. Once "shift by $\mu$, scale by $\sigma$" feels natural, the formula stops being something to memorise.

πŸŽ“ Practice Problem

Question. Suppose 80% of students hand in homework on time. From a random sample of $n$ students, $\Pr(\hat{P} > 0.85) = 0.1056$. Find $n$.

Guided steps

  • $\mu = 0.80,\ \sigma = \sqrt{\dfrac{0.16}{n}}$
  • $\Pr(\hat{P} > 0.85) = \Pr\!\left(Z > \dfrac{0.05}{\sqrt{0.16/n}}\right) = 0.1056$
  • Convert to a left-tail probability: $\Pr\!\left(Z < \dfrac{0.05}{\sqrt{0.16/n}}\right) = 1 - 0.1056 = 0.8944$
  • Use invNorm(0.8944, 0, 1) β†’ $z \approx 1.250$
  • Solve $\dfrac{0.05}{\sqrt{0.16/n}} = 1.25$ for $n$

Answer: $n = 100$

🌻 Three Lines to Take Away

1️⃣ Subtract $\mu$ β€” move the centre to $0$.

2️⃣ Divide by $\sigma$ β€” change the ruler to "standard deviations".

3️⃣ Same event, new coordinates β€” probability is preserved.

That is why $Z = \dfrac{X-\mu}{\sigma}$ is not a formula β€” it is the same story told in standard units.

β€” Dally 🌻 Β· for Emma's class

πŸ”Š Narration