STATISTICAL EXPERIMENT

Tags: #◇Computer Science #◇STT101A #◇Conceptual #◇Statistical Experiment

---

◇Overview

A Statistical (Random) Experiment is any well-defined process or observation whose outcome cannot be predicted with certainty before it is performed, even though the complete set of all possible outcomes (the sample space $S$) is known in advance.

◇Key Definitions

• Three Defining Characteristics of a Statistical Experiment:
  1. Well-Defined: The experimental procedure and the set of all possible elementary outcomes ($S$) are completely specified before the trial.
  2. Non-Deterministic (Uncertain Single Trial): On any individual trial, the exact outcome $\omega \in S$ cannot be predicted with certainty.
  3. Repeatable with Statistical Regularity: Under identical conditions, the experiment can (at least conceptually) be repeated, and long-run relative frequencies stabilize toward fixed probabilities.
• Deterministic vs. Statistical Experiment: Compiling a fixed source file with a deterministic compiler always yields the exact same binary (deterministic); measuring network packet latency across the Internet yields fluctuating values on each trial (statistical).
• Trial: A single execution or realization of a statistical experiment, producing exactly one outcome $\omega \in S$.

◇Worked Example

Problem: Classify each process below as a Deterministic or Statistical experiment:

1. Calculating $\sqrt{144}$ on a calculator $\to$ Deterministic (always $12$, no uncertainty).
2. Measuring the number of bit flips caused by cosmic rays in satellite DRAM over a 24-hour period $\to$ Statistical ($S = \{0, 1, 2, \dots\}$, uncertain single trial, follows a Poisson distribution).

◇References & Navigation

• Next Topic: ◇Sample Space
• Related Notes: ◇Event