Do outliers begin from or above the whisker-limit?

Question

Does outliers begin on the whisker limit or above it?

In the (Python) example below the calculcated upper whisker limit is 64.8125. Is a value of 64.8125 an outlier (>= upper_limit)? Or the outliers are higher then the limit (> upper_limit)?

>>> import seaborn as sns
>>> df = sns.load_dataset('titanic)
>>> df.age.quantile([.25, .5, .75])
0.25    20.125
0.50    28.000
0.75    38.000
Name: age, dtype: float64
>>> q1, q2, q3 = df.age.quantile([.25, .5, .75])
>>> iqr = q3 - q1
>>> whisker_length = iqr * 1.5
>>> upper_limit = q3 + whisker_length
>>> lower_limit = q1 - whisker_length
>>> lower_limit, upper_limit
(-6.6875, 64.8125)

Answer 1

Under Tukey's approach, the whiskers are specifically drawn in each direction to the furthest observation from the median that is still within (or equal to) the inner fence. If a point is exactly on the inner fence, the whisker includes it.

A marked point must therefore be outside the inner fences.

That is strictly above on the high side (above the upper inner fence) but strictly below on the low side (below the lower inner fence).

Here's an illustration where there's points above the upper inner fence.