Sample Design Paper Outline

STAT 522 – Sampling Design

Sample Design Paper Outline

Due Friday, April 13

In this project you will be designing a sample using at least one complex sampling feature (stratification, clustering, unequal probabilities of selection, etc.) and likely more. This should be a real example, where you investigate how you will obtain a sampling list, available information on the population including information for strata and clusters, and optimal allocation. The goal is to get the sample that is the best in terms of representing the population and minimizing sampling error.

Follow the outline below as a general guideline. Papers should be about 5 pages long with additional tables and graphs included where necessary, but could be shorter or longer depending on what amount of space you need to cover the topics in the outline below.

I. Define the population of inference and what you want to estimate about that population.

a. Describe the population including: what the population units/elements are, the real (measured) information you can obtain on that population especially information that you could use to define strata, clusters or that you might use in optimal allocation.

b. Describe what you want to estimate for that population in terms of the variable(s) and the type of estimate (means, totals, ratios, domain estimates, other). You should have at least one focal variable and possibly two. You should have at least one, but possibly 2 or 3 estimates.

II. Describe how you will take the sample.

a. Describe the sampling frame or list used to draw the sample. This may have to be a list of clusters if there is no good list of the population elements. The design of your sample will depend on the list you can acquire to some extent.

b. Describe the sample design including how you will randomize selection and what the probability of inclusion for each observation will be. This is where you will describe strata, clusters, one or multiple stage selection, allocation within strata, and sample sizes. Give the sample size(s) for all stages as well as the final sample size. Assume that you have limited financial and time resources. Estimate about what it will cost to collect the needed data on each of the elements in the sample. For example, the cost to do a phone survey or in-person observations. Where possible, use equations from the text to get optimal sample sizes using real or estimated variances and/or cost. Calculate the final probabilities of selection based on the sampling design.

III. Describe how you will analyze the sample and calculate design effects

a. Present the equations you will use to estimate your statistics including equations for the point estimates, the variance estimates (or standard errors), and the confidence intervals. The estimators will depend on whether you have clustering, strata, etc. You may also include post-stratification, ratio, or other estimators that are more precise.

b. Use equations from the text and class to calculate the design effects for each sampling feature separately (clustering, stratification, and weighting). I will help you obtain these.