Sampling frames Flashcards
What is a drawback of using telephone directories as a sampling frame?
Undercoverage is likely as this will not include people without a home phone, people who are ex-directory, people with no fixed address, people that have a mobile phone. Also, if you call a house who you speak to will affect the evidence gathered, especially if you are accepting by-proxy answers. This is known as ‘clustered’ as people essentially share the same listing on the sampling frame. Normally there is a random approach to decide who should be questioned but what if they are not in!
What is a dual sampling frame?
When you use 2 sampling frames to select your sample. For example, you could use census data to select places where we will do our study (LAs) and then local directories to select households from those areas.
Why are omissions on sampling frames such a problem?
The missing units tend to share common characteristics - like all people without a phone.
What is a method of mitigating missing phone numbers in a telephone directory?
- Random digit dialling which reaches listed and non-listed numbers.
- Using more than one sampling frame.
- Incorporating mobile phone numbers.
Are sampling frames made for statistical purpose?
Not usually so you can clean your sampling frame through accounting for missing units, units included that should not be and duplication. You should consider how the list was compiled, how and when deletions are made and who might be left out.
What is a sampling frame?
A list of all the population units within the study population. This needs to be up-to-date to ensure the sample is representative.
What are some examples of sampling frame?
- Postcode Address File (large users file and small users file)
- The Electoral Register
- The Inter-Departmental Business Register
- Telephone directory
- DfE schools data
- OS AddressBase
What are some common problems with sampling frames?
Undercoverage - people with no home, new housing estates, caravan sites, people who live on boats, people who are not registered to vote, people with no landline, ex-directory.
Overcoverage - people who have moved but are registered in 2 places, deaths, UK students and armed forces abroad, people who have 2 homes, people with 2 phone lines going into the house.
Why is it a problem if people in your sample refuse to take part?
This non-response is an issue as it it likely that the people that refuse have a similar profile to each other and will be missed out the survey. The findings will be biased toward those that were willing to participate.
What would be a drawback of using dual sampling frames for landline and mobile phone numbers?
Many people have both so you would need to weight the data to take into account that some units would have had a higher chance of selection.
What should you do if you know that some units had a higher chance of being selected and this was not an intentional part for the study to represent the population?
Weights are needed so that over-sampled groups are not over-represented in the statistics.
When using telephone surveys, what can you do if you want each person within households to have an equal chance of selection but there are different amounts of people living in the different households?
You can weight the data accordingly. The probability of selection times the weight should be the same for each respondent.
So weighting can account for the probability of selection being higher.
