Skip to content
Omnikalc

Neyman allocation calculator

Enter the size and standard deviation of three strata and the total sample size to get the optimal number of units to sample in each, the standard error and the gain over proportional allocation.

Updated on
Your inputs

Value out of range

Value out of range

Value out of range

Value out of range

Value out of range

Value out of range

Value out of range

Results

Units to sample in stratum 1

105.3

Units to sample in stratum 2

126.3

Units to sample in stratum 3

168.4

Standard error with Neyman allocation
0.9500
Standard error with proportional allocation
1.1068
Variance reduction against proportional allocation
26.3 %
Large gain: the strata differ a lot in spread
Proportional allocation, stratum 1
200.0
Proportional allocation, stratum 2
120.0
Proportional allocation, stratum 3
80.0
Any allocation larger than its stratum? (1 = yes)
0
All allocations fit within their strata

The Neyman allocation calculator splits a stratified sample of fixed size across three strata to get the smallest variance of the mean, and compares it with proportional allocation.

How the calculation works

  • Enter the total sample size.
  • Enter the size and the standard deviation of each stratum.
  • The calculator gives the units to sample in each stratum, the two standard errors, the gain and an overflow check.

Example

A population of 10,000 units has strata of 5,000, 3,000 and 2,000 with standard deviations of 10, 20 and 40. For a sample of 400, the sums N × S are 50,000, 60,000 and 80,000, so the allocation is 105.3, 126.3 and 168.4 units. The standard error is 0.9500, against 1.1068 with proportional allocation (200, 120, 80), a variance reduction of 26.3 %.

Why it matters

The calculator helps survey statisticians, auditors who sample accounts of different amounts, quality inspectors who sample batches and students of sampling theory.

Good to know

The standard deviations come from a pilot study, past surveys or an auxiliary variable, and a rough guess still works well because the allocation is not very sensitive to small errors. Round the allocations to whole numbers while keeping the total, and make sure no stratum gets more than its size. The standard errors ignore the finite population correction, which lowers them when the sample is a large share of the population. In practice, many surveys keep a minimum of two units per stratum so that the variance can be estimated within each. For a single estimate that matters a lot, check the gain again before changing a plan that is already running.

The sample size calculator sizes a simple random sample, the standard error calculator measures the precision of a mean, the design effect calculator quantifies the effect of the design and the systematic sample interval calculator picks every k-th unit.

Frequently asked questions

What is Neyman allocation?

It is the way of splitting a stratified sample that gives the smallest variance for the mean at a fixed total size. The sample in stratum h is proportional to N_h × S_h: bigger and more variable strata get more units.

How does it differ from proportional allocation?

Proportional allocation samples each stratum in proportion to its size only. Neyman allocation also accounts for variability, so it samples a small but highly variable stratum more heavily. When all the S_h are equal, the two coincide.

What if the costs differ between strata?

The optimal allocation then also divides by the square root of the cost per unit: n_h is proportional to N_h × S_h / √c_h. This calculator assumes equal costs.

Was this calculator helpful?

Thanks for your feedback!

Results are estimates provided for information purposes. They do not replace professional advice.