Population vs Sample: The Glimpse Problem

What You Want vs What You Can See

The population is the complete universe of all possible data — every word ever spoken, every photo ever taken, every customer who could ever exist. You can never observe the whole population. The complete set is too big, too distributed, often partly in the future.

The sample is what you actually observe — a finite subset. Your training dataset. A poll of 1,000 voters. Last quarter's customer behavior. Your goal as a statistician (or ML engineer) is to use the sample to infer something true about the population.

Sample → Population Bridge

The whole game of statistics is figuring out how much you can trust an inference from sample to population. A bigger sample buys you more confidence; a biased sample buys you wrong answers no matter how big.

In ML: training data is a sample. The deployed model has to generalize to a population that includes inputs it never saw. The closer your sample resembles the population, the better the model generalizes. Data quality is sample quality.

Population is the truth. Sample is what you have. The art of ML is making the leap responsibly — and knowing the leap is always a bet.

Code

Estimation as sample size grows·python

import numpy as np

# True population mean (which you don't know in real life)
population = np.random.normal(loc=100, scale=15, size=1_000_000)
true_mean = population.mean()
print(f"true mean: {true_mean:.3f}")

# Take a sample of 30 — what do we estimate?
sample = np.random.choice(population, size=30)
print(f"sample mean (n=30): {sample.mean():.3f}")

# Take a bigger sample
sample = np.random.choice(population, size=10_000)
print(f"sample mean (n=10000): {sample.mean():.3f}")
# Bigger samples → estimates converge to the truth

Exercise

Generate a population of 1,000,000 random numbers. Take samples of size 10, 100, 1000, 10000. For each, compute the sample mean. Plot or print all four. Notice how the sample mean wobbles for tiny samples and stabilizes for big ones.

Hint

np.random.choice(population, size=n) gives you a sample. Bigger n → estimate closer to the true mean — that's the law of large numbers showing up.

Population vs Sample: The Glimpse Problem

What You Want vs What You Can See

Sample → Population Bridge

Code

External links

Exercise

Progress

Comments 2