Which sample-size method should I use?

It depends on the statistic used to test your hypotheses. If the population is known and you use basic statistics, use Taro Yamane or the Krejcie & Morgan table. For multiple regression or SEM, G*Power is now preferred because it is based on statistical power and effect size.

How is Taro Yamane different from Krejcie & Morgan?

Both are used when the population (N) is known. Taro Yamane is a formula where you set the margin of error (e) yourself, while Krejcie & Morgan is a ready-made table you read directly at 95% confidence and e = 0.05.

Why do some advisors no longer accept Taro Yamane?

Because it is based on population size alone and ignores statistical power and effect size. For studies using advanced statistics, many universities now require G*Power instead.

Taro Yamane vs Krejcie & Morgan vs Cochran vs G*Power: Which Sample-Size Method?

"How many respondents do I need?" trips up many Master's students because there are so many formulas to choose from. Here is a 5-minute summary of the four most popular methods and how to pick one.

Overview of the four methods

Method	Use when	You need	Strength
Taro Yamane	Population is known	N, e	Simple, adjustable e
Krejcie & Morgan	Population is known	N (read table)	Fast, no calculation
Cochran	Large/unknown population	p, q, e, Z	Flexible, infinite population
*GPower**	Advanced stats (regression, SEM)	Power, effect size, α	Highest statistical rigor

1. Taro Yamane (1967)

Best when you know the exact population and use descriptive or basic tests. Formula: n = N / (1 + Ne²). Its weakness is ignoring statistical power, so it is losing acceptance for complex designs.

2. Krejcie & Morgan (1970)

Uses similar statistical logic but delivered as a ready-made table at 95% confidence. Just look up your population N and read off the required n.

3. Cochran (1977)

The most flexible when the population is unknown or very large, using proportion (p, q), confidence level (Z), and margin of error (e), with a finite-population correction available.

4. G*Power (today's standard)

If your study uses multiple regression or SEM, this is the standard leading universities require, because it is computed from statistical power, effect size (f²), and significance level (α).

So which should you choose?

Survey research, known population → Taro Yamane or Krejcie & Morgan
Unknown / very large population → Cochran
Regression, path analysis, SEM → G*Power

Try our free G*Power sample-size calculator →