Systematic and unequal sampling, HT-estimator

19 important questions on Systematic and unequal sampling, HT-estimator

What is a serious bias that can occur in quota sampling?

Selection bias: the most accessible members are likely to differ from less accessible.

Why can estimates from quota sampling be biased (compared to probability sampling)?

Estimates from quota sampling can be biased because it is not a probability sample. Therefore, not all population element has a non-zero inclusion probability and thus estimates can be biased.

Can systematic sampling give a representative sample of the dataset?

Yes, when the sampling frame is a random order, systematic sampling can give a representative sample of the dataset. However, when the list is not ordered randomly, the systematic sample is biased.
  • Higher grades + faster learning
  • Never study anything twice
  • 100% sure, 100% understanding
Discover Study Smart

Why is a replicated systematic sample used?

A replicated systematic sample is used because it can be difficult to compute the SE of the estimator (uncertainty) with systematic sampling, and this gives problems with analyzing the data. Replicated systematic sampling is thus basically a trick to compute the variance of the estimate.

How does a replicated systematic sample work?

1. In a replicated systematic sample, a proportion of the original sample is taken r times, such that rF = n.

F is the size of the zone.
rF is the length of the interval with which the systematic sample is taken.

Then, the sampling fraction is 1/rF    
The length of each replication is n / r = N / rF

Sample n = N/F
Sampling fraction f = 1/F = r/rF

2. For each of r replications, compute the statistic.
These statistics are combined to compute the final sample statistic. 
(The Horvitz-Thompson estimator cannot be used to compute the variance estimation! )

What is required for systematic replicated sampling?

It requires r independent samples that are similar in design.

Why can an SRS have a higher standard error than a systematic sample?

Because with an SRS there are more possibilities. Systematic sampling gives a more accurate representative sample when the list is in random order.

What is an advantage of unequal probability sampling?

With the proper choice of selection procedure, estimators are much more precise. The variance is smaller as the selection probabilities of the elements are more proportional to values of the target variable.

How can a sample with unequal probabilities being drawn?

Drawing a sample with unequal probabilities is realized in practice by looking for an auxiliary variable that has a strong correlation with the target variable. All values of the auxiliary variable must be positive.

What are four reasons for post stratification weighting?

1. Variance reduction,
2. Under-coverage,
3. Unit non-response. 
4. To adjust the weighted sample distribution for key variables of interest to make it conform to a known population distribution.

What does a sampling weight represent?

The sampling weight represents the number of units in the population that the respondent represents.

How is a sample weight computed?

A sample weight is the inverse of the inclusion probability of a element, thus: 1 / inclusion probability.

What is the Horvitz-Thompson estimator?

The Horvitz-Thompson estimator is an estimator in which the sampling weights are taken into account in the calculation of the estimate.

When is the Horvitz-Thompson estimator particularly useful?

The Horvitz-Thompson estimator is especially useful in complex sampling, when the inclusion probability can vary by each respondent, depending on the sampling scheme.

What is a self-weighting sample?

In a self-weighting sample the weights of all sampled units are the same.

What are three reasons why samples are rarely self-weighting?

1. Sampling units are selected with unequal probabilities of selection.
2. The selected sample often has deficiencies including non-response and non-coverage.
3. The need for precise estimates for domains and special subpopulations often requires oversampling these domains.

What are three possibilities for estimation for unequal probability sampling?

1. The Horvitz-Thompson estimator can be used (but variance estimation is not possible).
2. The estimate can be estimated with the use of an augmentation variable, say X.
3. Use sampling with replacement.

For which kind of estimations do sample elements need to be weighted by the reciprocal of their selection probabilities and for which not?

For estimating totals, sampled elements need to be weighted by the reciprocal of their selection probabilities.

For estimating means and proportions, the weights need only be proportional to the reciprocals of the selection probabilities.

How can a non-response adjusted weight be defined?

wh = w1h * w2h = w1h * nh / rh

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