Basic concepts of sampling (stratified sampling)

15 important questions on Basic concepts of sampling (stratified sampling)

How is an estimate variance constructed for stratified sampling?

Stratum variances are added into a combined estimate.

What are 6 reasons for stratified sampling?

1. The strata are domains of study or interest.
2. Avoid possibility of a really bad sample (mixed population of males and females).
3. The population is geographically diverse, stratification is used to organize the sample and data collection.
4. Sample data of known precision for subgroups.
5. Stratified sampling may lower the costs of the survey.
6. Stratified sampling often gives more accurate estimates of population means and totals.

What are two benefits of stratified sampling?

1. Improved representativity.
2. Improved precision of estimates, specifically if strata are chosen such that they are homogeneous with respect to the target variable.
  • Higher grades + faster learning
  • Never study anything twice
  • 100% sure, 100% understanding
Discover Study Smart

What are the risks of over- and understratification?

Unnecessary (over-) stratification can degrade precision. Understratification risks bias.

What is a design effect?

A design effect is a ratio of variances and can be computed to compare one sampling design with another. It can be used to calculate how many observations are needed to get the same accuracy with a sample design that is more precise.

E.g. var(strat mean)/var(srs mean) = 0.83
0.83*300 = 250 observations to obtain same accuracy.

Are the mean and variance estimator of a stratified sample biased?

No, the mean and variance estimator of a stratified sample are unbiased.

What are requirements for stratification variables?

1. Stratum sizes must be known.
2. It must be possible to select a sample in each stratum separately.

Stratification variables should be variables that have a strong relationship with the target variable (then, the variances in the strata are small).

What methods can be used to determine which grouping is the most effective (smallest variance) in the case of a quantitative grouping variable?

The cumulative-square-root-f rule:
1. Frequency distribution
2. Square root of the product of the frequency (f) and the interval width (w) is computed for each value of the variable.
3. Values are grouped in such a way that the sum of the computed quantities is approximately the same in each group.

What is proportionate to size sampling?

When you sample in a stratified sample proportionate to size, you want an equal sampling fraction in each stratum. Thus, in each sample, the same proportion of elements are sampled, while the count can be different.

Another view at proportionate to size sampling is that you make sure that the population size of the stratum divided by the total population size is equal to the sample size of the stratum divided by the total sample size:

Nh / N = nh / n

What happens to the weights of the strata when you sample proportionate to size?

Proportionate to size sampling leads to a self-weighting sample: fh = f.

What is optimal allocation?

Optimal allocation is a division of the sample size over the stratums to achieve the least variance for the overall mean, while taking costs into account.

What is a consequence of optimal allocation for the inclusion probabilities?

The inclusion probabilities of the elements are equal to nh / Nh. Following from the formula for optimal allocation, the inclusion probabilities are all proportional to the stratum standard deviations. As a consequence, not every element has the same inclusion probability. This is no problem, because the estimator of the stratified sample mean corrects for this.

When is stratification inefficient?

Stratification is inefficient when there is only a small gain in precision. Stratification complicates the analysis and is then not worth the small gain in precision.

When is stratification efficient?

Stratification is efficient when stratum means differ widely (homogeneity within strata and heterogeneity among strata).

Why does homogenous strata lead to a more precise estimator?

In the case of homogenous strata, variation is due to between strata means, not within strata means, this heightens the precision of the estimator.

The question on the page originate from the summary of the following study material:

  • A unique study and practice tool
  • Never study anything twice again
  • Get the grades you hope for
  • 100% sure, 100% understanding
Remember faster, study better. Scientifically proven.
Trustpilot Logo