Modified CASPERs

Increase Cluster Selection

If you are worried one or more clusters may be inaccessible because of damage or restricted entries, you may consider increasing the number of clusters selected a priori. While tempting to select a substitution cluster once in the field, this is not allowed because it negatively affects the representative of the data. If you are concerned about inaccessibility, you have two options:

1. Exclude the area from the sampling frame before selecting the 30 clusters (keeping the traditional CASPER methodology). NOTE: the results will not represent those excluded areas)
2. Increase the number of clusters selected prior to going into the field (this is a modified CASPER)

Any a priori change to the number of clusters selected (i.e., any number other than 30) is considered a modified CASPER and should be noted as such in the methodology section and throughout your report(s). For example, you may decide to select 35 clusters instead of the standard 30. If this is chosen, it is essential that teams visit all 35 clusters and treat the design as a 35×7 (goal of 245 interviews), potentially requiring more time and resources to complete. It is important to remember that oversampling does not improve response rates but can increase the sample size (and thus power). For more information, please see the CASPER toolkit, Section 2.5.1pdf icon.

Low Number of Households

A problem may arise, especially in more rural areas, that Census-based clusters may have few households, making it difficult or impossible for teams to complete the needed interviews from that cluster. Generally, this is not a major issue as smaller clusters have lower probability of being selected and therefore those with fewer than seven households will be kept to a minimum. However, if the sampling frame consists of a relatively high proportion of small clusters (i.e., fewer than 10-15 households), interview teams may have difficulty finding seven households to interview resulting in a low completion rate. If you are concerned that this may be a problem, there are three options:

1. Use the larger Census block group variable and keep the traditional CASPER 30×7 design. Note: Census block groups may be geographically too large to be manageable, especially in rural areas
2. Adjoin census blocks together using GIS software to create larger clusters and keep the traditional CASPER 30×7 design
3. Conduct a new sampling methodology, such as stratified sampling or systematic sampling, when the first two options (block groups and combining clusters) make clusters too geographically large or limit the total number of clusters to where cluster sampling is not appropriate. NOTE: This is no longer considered a CASPER or modified CASPER and must not be described as such in any report (an option would be to describe as a “community assessment” or “rapid needs assessment”)

For more information, please see the CASPER toolkit, Section 2.4.1pdf icon and 2.5.2pdf icon or contact Amy Helene Schnall at GHU5@cdc.gov for cluster combination assistance

Urban and Rural Sampling Frame

Many parts of the United States comprise both rural and urban areas. Because a traditional CASPER selects clusters probability proportional to size, it is likely that most selected clusters will be from urban areas with higher household counts, leaving few selected clusters form the rural areas of the sampling frame. If you desire representation from both urban and rural areas, there are two options:

1. Conduct two separate CASPERs: one in the urban area, one in the rural area (keeping the traditional CASPER 30×7 design in each).
2. Conduct a three-stage sampling design where you would first stratify the sampling frame into “urban” and “rural” strata (or sections) then select clusters from each strata (for example 15 in each) to assure both areas are represented. NOTE: You will need to account for the third stage in the weighting. This is considered a modified CASPER as it still retains the clusters sampling and should be noted as such in the methodology section and throughout the report(s).

For more information, please see the CASPER toolkit, Section 2.5.3pdf icon

Individual-Level Questions

CASPER is designed to be household-based and, therefore, all questions should be asked at the household level. If individual-level questions are necessary, consider whether CASPER is the best method to reach your objectives. If the majority (i.e. 90% or more) of your questionnaire is household-based but there are some questions that must be asked at the individual level (e.g., mental health questions such as the Generalized Anxiety Disorder [GAD-2] or Patient Health Questionnaire [PHQ-2]), it is acceptable to ask a limited number of questions at the individual level. To do so, you must take the following steps:

1. Include a question to quantify the number of adults living in the household.
2. Randomly choose an adult in the selected household who will respond to the individual-level questions.
3. Calculate a separate weight during the analysis for the individual-level questions.

Individual weights are described in the CASPER FAQs. You can also contact CASPER@cdc.gov for guidance and more information.

This is considered a modified CASPER and should be noted in the methodology section and throughout your report(s).