The recommended sampling process for countries that will select their ultimate sample of MPNs in Option 1 is somewhat different than for those countries implementing Option 2. Guidelines for all these scenarios are presented below.
Countries Selecting MPNs
Countries that plan to select MPNs using Option 1 should follow the following guidelines:
- MPNs should be generated in volumes that can be reasonably managed by the computing and network capacity.
- Because the number of MPNs selected as specified by the sample design will likely vary by a considerable amount, MPNs should be selected via SRS using RDD without replacement for Option 1 and via SRS without replacement for Option 2 to provide a simple random sample. This variation in the number of MPNs selected can be due to many possible sources, such as MP market penetration, aversion to surveys by the public, or percent of possible MPNs actually assigned.
- Methods can be used to improve the sample of MPNs with RDD in Option 1, including number validation. These methods increase the proportion of working MPNs in the sample by screening out invalid or inactive numbers prior to fieldwork. An initial sample of MPNs can be drawn from a sampling frame of randomly generated MPNs that is first screened to exclude banks reserved for operator technical services. These numbers can be further screened for voicemail and operator messages to be classified as unknown eligibility if eligibility status cannot be determined.
- It is important that every MPN on the sampling frame has some nonzero probability of selection and that the probabilities of selection are retained in the final analysis file. These probabilities of selection are one of several pieces of statistical information that are used to compute sample weights for the ultimate respondents to the NCD Mobile Phone Survey.