sample

A sample is a specific kind of data — a subset of observations drawn from a larger population for analysis. Per Law 9, a competent speaker would call a sample 'a type of data' used for representational purposes.

Random sample IS a specific kind of sample — it is a sample selected with equal and known probability for each member. A competent speaker would call a random sample 'a sample.' The INSTANCE_OF arrow and nearest-kind filing are correct: sample is the general kind, random-sample is the specific.

Present-tense necessity (Law 8b): the empirical distribution is a functional ON the sample — F_n(x) = (1/n)·Σ 1{X_i ≤ x} is defined over the sample values, and the entry's own carve says it 'describes observed data values from a finite sample.' Remove the sample now and the functional has no domain to operate on: there is nothing to describe, no F_n to compute — it stops OPERATING, it is not merely sayable. Not an ambient dependency (contrast: X DEPENDS_ON mathematics = STRIKE): the sample is the specific operational input, the same functional→input pattern as variance→probability-distribution and MLE→likelihood. (Note: the 'data' entry carves data CLEANING, so it is not the correct substrate; the carve's named input is the sample.)