Source
Adapted from my own handwritten and hand-typed notes from Year 1 to Year 3 of my EEE degree
Many measured variables have a roughly bell shaped normal distribution, fully described by its mean and standard deviation .
Z-Score
The z-score of a data point is how many standard deviations it lies from the mean:
Positive means above the mean, negative means below.
Empirical Rule
| Interval | Share of data |
|---|---|
| 68.27% | |
| 95.45% | |
| 99.73% |
Split by band: 34.14% between and , 13.59% between and , 2.14% between and (same on the left side).
Confidence Intervals
A confidence level (like 95%) says how sure we are that a value falls inside a range around the mean:
| Confidence | Interval |
|---|---|
| 90% | |
| 95% | |
| 99% |
Example: Pokemon HP
Taking and :
| Mark | |||||
|---|---|---|---|---|---|
| HP | 18.25 | 43.75 | 69.25 | 94.75 | 120.25 |
So if HP were normal, about 68% of Pokemon would have HP between 43.75 and 94.75. A Pokemon with HP 145.75 has , which would be very rare.
from scipy import stats
z = stats.zscore(pkmndata['HP'])
(abs(z) <= 1).mean() # actual share within 1 SDThe z-score builds directly on the mean and SD in Univariate Statistics.