Skip to content
Omnikalc

Ratio estimator calculator (survey sampling)

Enter the sample means of the variable of interest and of an auxiliary variable, the known population total of the auxiliary variable, the standard deviations, their correlation and the sample and population sizes to get the ratio estimate of the total, its standard error and the gain over simple expansion.

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

Value out of range

Results

Ratio estimate of the total

26,541.7

Standard error of the ratio estimate
496.7
Estimated ratio
1.0833
Simple expansion estimate of the total
26,000.0
Standard error of simple expansion
909.9
Variance ratio (simple over ratio)
3.36 ×
Strong gain: the auxiliary variable helps a lot
Correlation needed for any gain
0.497

The ratio estimator calculator estimates a population total using a known auxiliary total, with its standard error, and compares it with simple expansion to show the gain in precision.

How the calculation works

  • Enter the sample and population sizes and the sample means of the two variables.
  • Enter the known auxiliary total, the standard deviations and their correlation.
  • The calculator gives the ratio, the ratio estimate, the simple estimate, the standard errors and the gain.

Example

A sample of 40 from 500 shops has a mean of 52 in the current year and 48 for last year, whose population total is 24,500. The ratio is 52 / 48 = 1.0833, so the ratio estimate of the total is 1.0833 × 24,500 = 26,541.7. With standard deviations of 12 and 11 and a correlation of 0.85, its standard error is 496.7, against 909.9 for simple expansion (500 × 52 = 26,000.0). The variance is 3.36 times smaller.

Why it matters

The calculator helps students of survey sampling, official statisticians who use administrative data, auditors who estimate totals from a sample, and analysts who use last year’s figures to sharpen this year’s estimate.

Good to know

The approach pays off when Y and X are strongly and proportionally related, with a line through the origin; if the relation has a large intercept, a regression estimator is better. The standard deviations and the correlation come from the sample. The auxiliary total must be known for the whole population and be measured in the same way as in the sample. With small samples, below about 30 units, the large-sample variance formula can be optimistic. The finite population correction (1 − n/N) is included in the variance, and it matters when the sample is a large share of the population.

The Neyman allocation calculator optimizes stratified samples, the standard error calculator measures precision, the sample size calculator sizes simple samples and the design effect calculator compares designs.

Frequently asked questions

What is a ratio estimator?

When the variable of interest Y is roughly proportional to an auxiliary variable X whose population total is known, the ratio R = ȳ / x̄ from the sample is applied to the known total: the estimate of the total is R × X. It corrects the sample for being too high or too low in X.

When is it better than simple expansion?

When the correlation between Y and X is high enough, roughly above R × S_x / (2 × S_y). A strong proportional relationship cuts the variance a lot; a weak one makes the ratio estimator worse. Typical auxiliary variables are last year’s value, the size of a unit or an administrative count.

Is it unbiased?

Slightly biased in small samples, but the bias shrinks quickly as the sample grows and is usually small compared with the standard error. The variance formula used here is the usual large-sample approximation.

Was this calculator helpful?

Thanks for your feedback!

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