survey and sampling Flashcards
what are surveys?
a set of simple questions that give us an idea of whats happening commonly in the form of questionaires and can be online or written
it takes a subset of the population
how do you find a random sample?
mean/ x and x is assumed to be 50
have to be 95% confident the mean is between 40 to 60
method 1 of finding the random sample
first find your range max
take out numbers using their first few numbers of digits depending on max range and only numbers below the range can be noted
you cant include the same number twice
after his find each number corresponding to each number youve found witht in the range
add them all up and divifr by how many numbers
method 2 of random sample
when can you overprepresent in sampling?
in smaller groups there may be a slight overrepresentation
how do you figure out how many people to select?
first find the percentage of the entire population for each category
once found do the percent to the sample you will use and that will give you how many should be selected for each part of the survey
Simple Random Sampling
Type of Probability Sampling
Sampling units have same probability of selection
Need a sampling frame
Complete list of population
Stratified Random Sampling
Population might be made up of different strata, e.g. gender, ethnicity, …
Simple random sampling may miss out or over-represent certain groups
Divide population into groups
Take simple random sample from each group
Sample frame required
Including information pertaining to each stratum
Systematic Sampling
Type of probability sampling method
sample selected according to a random starting point but with a fixed interval.
Sampling interval, is calculated by dividing the population size by the desired sample size.
Choose first sampling unit at random
Subsequently select every nth sampling unit
Systematic Sampling: Problems
Not a true random sample
Can lead to bias – e.g. if every nth sampling unit is similar and not representative
example of systemic sampling
Example: systematic sampling from a list numbered 1 to 120
Population size is 120 and required sample size is 10
120 / 10 ≈ 12 so use an interval of 12.
Pick a reference number between 1 and 12 at random
Suppose number picked is 10
ID numbers selected will be:
10, 22, 34, 46, 58, 70, 82, 94, 106, 118.
The sample will have similar properties to a simple random sample if the list you selected from is randomly ordered.
Possible practical problems with sampling
Questionnaire not completed / not returned
Cluster Sampling
A cluster can be considered a representative sample which is, conveniently, located in just one place - “ready made”.
A cluster must be representative of the population in that
the “mix” of people within the cluster is similar to the mix in the population. We survey everyone in the cluster.
A good cluster will be similar to the final (overall) stratified random sample but without all the hard work!
Examples of cluster sampling
Population: 1st year students Cluster: lecture group
Population: bank customers Cluster: street residents
which sampling technques Requires the population list?
Cluster Sampling NO
Stratified Random Sampling YES
Quota Sampling NO
Simple Random Sampling YES
Systematic Sampling YES
How to do Quota sampling
Decide which groupings of people are relevant to your survey
Suppose you decide on age and gender.
Assuming you have no list of the population, base age and gender proportions on information about the population of UK.
Calculate the quota size for each age and gender combination from these proportions and your overall sample size.
Stop passers by and interview them, if they allow this.
Vary time and location. Stop when all quotas are filled.
Data Collection Methods
Postal/Online Questionnaire
Telephone Interview
Personal Visit
Street Interview
Positivies off Postal/Online Questionnaire
Relatively cheap
Easy to organise
Potential issues with postal/Online Questionnaire
No opportunity to clarify
Non-response bias – send reminders, offer incentives
Low response rates – as above
Positives of telephone Interviews
Cheap and easy to organize (skilled interviewers can complete a lot of surveys in a day of work.)
large audience for gathering a representative sample to complete the survey.
High response rate than web-based survey
Clarify questions
Potential issues of personal Visits
Interviewers need training
Expensive
Potential issues of telephone Interviews
Can annoy people – arrange call time before if possible
Time consuming
Requires training for interviewers
Positives of personal Visits
High response rate
Clarification of ambiguities
Good for collecting detailed information
Positives of street Interviews
High response rate
Clarification of ambiguities
Often used with quota sampling
Potential issues with street Interviews
Interviewers need training
Difficult to identify sampling units
Past QMB(A) Exam Question: In a large university offering a wide variety of degree programmes, an increasing number of students have been expressing concerns about coping with the maths content of their courses. The university’s learning support centre has designed a questionnaire to gather data on whether students are aware of existing facilities for help with maths and what the opinions are of those students who have used the facilities. The learning support centre only has sufficient resources to study a sample of 400 students.
a) Define relevant Population
b) Describe appropriate sampling method: pay attention to feasibility and potential issues
c) Select and justify most effective data collection method
a) All the students at Loughborough University whose degree has some maths content and who therefore might need to use the maths support centre.
b) Since LU will have a list of all its students and information on their degree programme, we can use a random sample selection method thus avoiding selection bias. We have the information we need to use stratified random sampling. To stratify, we can check the number of students on each degree programme with some maths content and then calculate the proportion in the population we have defined. We can then use the same proportions for the sample of 400 students. This will ensure that no degree programme is either over or under represented in our sample.
A problem with stratified random sampling is that not all students that are selected and asked to complete the questionnaire will respond. This will introduce non-response bias (for example students with “neutral” opinions or students from some degrees could be under represented). To encourage a good response rate, we can provide an incentive and send reminders to non-responders. Alternatively, we could use quota sampling but, because this is a non-random selection method ,we may get selection bias (students with no strong feelings about the maths centre may not want to take part in the survey).
c) We have the email addresses of all students at Loughborough university, hence we can send online questionnaires to the selected students. This is a cheap and effective approach as all students will be checking their email regularly, and is not more costly to send reminders.
Questionnaire Design
Must be clear about purpose of study before designing questionnaire
Otherwise, how do you know what to ask?
What do you need to know?
What question(s) do you need to get this information?
Designing an opinion survey: decisions
Aims: information needed is opinions about……..?
Given these aims, relevant population to survey is…..?
Survey the whole population or select a sample?
Best way to select a representative sample is……?
How can we best measure opinions about……?
Aspects of good questionnaire design
A good questionnaire will:
ask the right questions
be clear and unambiguous (but not offend)
be easy to complete
be readily transferred to a computer
To achieve this consider:
Content: what do I need to ask questions about?
Language: Simple? Brief? Clear? Respectful?
Structure: What will be a logical sequence of questions?
Response format: Appropriate for question? Clear?
Layout: Looks easy to follow and complete?
