Question

How to calculate emergency department wait time across the state if you don't have access to all the records. Which sampling technique would be the most accurate to use for this study ? Detail your reasoning of why did you choose this method.

Answer 1

To be sure for the context of the question, the inability of getting access to every department would either point out to following of the two cases:

1. When the out of all the records of the state emergency department, only some could be accessed (A Bigger Approach)

This is a typical case where application of principle of randomization could be confirmed. As the population units (records of all state emergency wait time records) cannot be accessed, sampling should be done. Based on the sample being gained, efficient statistics and estimators could be known, so that reliable inference about population mean wait time could be approximated.

As far as the type of sampling scheme to be applied is confirmed, Simple Random Sampling would beat every other sampling scheme like Stratified Sampling (Non Availability of all records within each strata), Systematic Sampling (Non Availability of population records to facilitate administration), Circular Sampling (it is a particular case of Systematic Sampling), and Ratio Methods of Estimation (due to absence of auxiliary variable). A simple random sample would enable random selection of units from the population without any bias and judgmental factors being included.

In case, area sampling is done, where within states, internally heterogenous and mutually homogenuos strata known as clusters (area sections say districts or blocks) could be formed, cluster sampling would be the best. In this scenario, each of these strata would be a realisation of the population and randomly selecting a few of these would be efficient to derive estimators. After clusters are selected, one stage sampling or multi stage sampling could be further applied.

2. When the records of some particular area (say districts or blocks) emergency departments cannot be be accessed (A Specific Approach).

In this case, it is on the researcher to create clusters which are representative of the population, keeping in mind all the population units of different kind are exhaustively included in these clusters. Then application of cluster sampling would be a better option, as the complete list of all population units is not available.