← Probability methods
Estimating an unknown σ (or μ) from a given percentile
The reverse of a standard z-score lookup: given that a certain PERCENTAGE lies below/above a value, work BACKWARDS to find the missing standard deviation or mean.The method
- Use inverse normal on the GIVEN percentage to find the z-score for that cutoff — a percentage below 50% gives a NEGATIVE z, above 50% gives a POSITIVE z.
- Substitute the z-score, the given x-value, and whichever of μ/σ IS known into z = (x − μ)/σ.
- Rearrange to solve for the UNKNOWN parameter (σ or μ) algebraically — this is a linear equation once z is known.
- Sanity-check the answer's sign and size against the context (a standard deviation must be positive; a mean should sit within the plausible range of the data).
Using the given percentage directly as the cumulative probability without checking above/below
Feeding 0.12 straight into inverse normal, when the question actually describes the percentage ABOVE a value, gives entirely the wrong z-score (and the wrong sign). Always convert an "above" percentage to its complementary "below" percentage (100% − given%) before finding z.| Given | Convert to (before inverse normal) |
|---|---|
| X% below a value | Use X% directly |
| X% above a value | (100 − X)% below |
| X% between two values | Combine with a known tail to isolate one cumulative probability |
Practice for this standard covers the whole topic — flashcards and quiz.