Inverse Normal Calculator

The inverse normal distribution is the quantile function of the normal distribution: it returns the value x below which a given fraction p of the distribution lies. It is written Φ⁻¹(p) for the standard normal and is also called the probit function or the normal percent point function.

The normal cumulative distribution function Φ(z) maps a z-score to an area between 0 and 1. The inverse normal reverses that map. Graphically, Φ⁻¹ is the reflection of the S-shaped CDF across the line y = x.

Important properties:

1. Domain: 0 < p < 1. The output covers all real numbers.
2. Symmetry: Φ⁻¹(1 − p) = −Φ⁻¹(p), so the 5th and 95th percentiles are mirror images.
3. Median: Φ⁻¹(0.5) = 0, so x = μ for p = 0.5.
4. No closed form: it cannot be written with elementary functions, so calculators use rational approximations such as AS 241.

To calculate the inverse normal, find the z-score with Φ⁻¹(p) and convert it with x = μ + z·σ.

1. Turn the area into a left-tail area. For a right tail use 1 − p. For a central area use (1 + p)/2. For two tails use 1 − p/2.
2. Find z. Look up the left-tail area inside a z-table and read off z, or use invNorm, NORM.S.INV, or this calculator.
3. Scale and shift. Multiply z by the standard deviation and add the mean: x = μ + zσ. A negative z gives a value below the mean.

Example: the top 10% of scores on a test with μ = 70 and σ = 8. The left-tail area is 0.90, z = Φ⁻¹(0.90) = 1.2816, and x = 70 + 1.2816 × 8 = 80.25.

The same probability gives four different answers depending on where the area sits under the curve. Mixing them up is the most common inverse normal mistake on exams.

For p = 0.05 with the standard normal:

• Left tail 0.05: z = −1.645, the 5th percentile.
• Right tail 0.05: z = +1.645, the 95th percentile.
• Central 0.05: z = ±0.0627, a narrow band around the mean.
• Two tails 0.05: z = ±1.960, the familiar 95% confidence bound.

Before calculating, sketch the curve and shade the region the question describes. If the shaded region touches only one end, it is a tail area; if it straddles the mean, it is central.

About 68% of a normal distribution lies within 1 standard deviation of the mean, 95% within 2, and 99.7% within 3. These are the central areas for z = 1, 2, and 3.

The rule is a quick estimate. The inverse normal gives the exact bounds instead:

• Exactly 68%: z = ±0.9945, slightly less than 1.
• Exactly 95%: z = ±1.9600, slightly less than 2.
• Exactly 99.7%: z = ±2.9677, slightly less than 3.

Because the rule rounds, use the calculator whenever a question asks for a precise cutoff, and use the rule to sanity-check the answer.

Every statistics tool has an inverse normal function, but each names it differently and some assume a left-tail area.

• TI-83/84: invNorm(area, μ, σ) in the DISTR menu, with an optional LEFT/RIGHT/CENTER tail setting on newer models.
• Excel and Google Sheets: NORM.INV(p, μ, σ) for any normal, NORM.S.INV(p) for z-scores. Both take a left-tail area.
• Python: scipy.stats.norm.ppf(p, loc=μ, scale=σ), or statistics.NormalDist(μ, σ).inv_cdf(p) with no extra libraries.
• R: qnorm(p, mean = μ, sd = σ), with lower.tail = FALSE for right tails.

If your software has no tail option, convert first: a right tail p becomes 1 − p, and a central area p becomes (1 + p)/2 for the upper bound.

A percentile cutoff is a left-tail inverse normal: the k-th percentile is μ + Φ⁻¹(k/100)·σ. This is how admissions offices set score thresholds, how pediatric growth charts mark the 3rd and 97th percentiles, and how manufacturers set tolerance limits.

• IQ (μ = 100, σ = 15): the 98th percentile, a common giftedness threshold, is about 130.8.
• SAT total (μ ≈ 1050, σ ≈ 210): the 90th percentile is roughly 1319.
• Adult height (μ = 175 cm, σ = 7 cm): the 5th to 95th percentile range is 163.5 to 186.5 cm, which designers use for doorways and seats.

Real data are only approximately normal, so treat percentiles far in the tails (below 1% or above 99%) as estimates.