5 Sampling And Data Collection

Question 1

Sampling

Answer 1

More beneficial than census because: budget and time restraints, accurate and reliable results

Sampling frame: complete list of all eligible contacts that can take part in research and represent target population

Question 2

Confidence interval approach

Sampling frame

Answer 2

Based on statistical estimations that take into account the margin of error, expectation about the prevalence of an event and level of confidence (z score) that results are reliable.

5% margin error acceptable.

Question 3

Probability v non probability sampling

Answer 3

Probability: every member of population has known and equal chance of being chosen. Selection based on chance and error related to researcher judgement in selecting respondents is eliminated. Population list available. Standard error and confidence intervals can be calculated. Generalisations can be made.

Non probability: every member of population doesn’t have known and equal chance of being chosen. Population list unavailable. Sampling error not known. Generalisations limited.

Question 4

Probability sample

Answer 4

Simple random sample- sampling procedure that ensures each element of population will have = chance of being included in sample. E.g. drawing names from hat.

Systematic sampling- sampling interval drawn from sampling frame.

Stratified sampling- samples are drawn reflecting proportions in existing population. Useful when population has skewed proportions.

Cluster- segmenting into geographic areas they’re expected to be representative of a heterogeneity in a population.

Question 5

Non probability sampling

Answer 5

Convenience sampling- obtaining units that are most conveniently available. Lecturer who uses students for a study. Avoid having people from same household.

Judgement sampling- experienced individual selects sample based on personal judgement about appropriate characteristic of sample member. Key accounts perceived to be capable of providing useful info.

Quota sampling- ensures various subgroups of population will be represented to the extent the investigator desires. (M/F 50/50). May not be representative.

Snowball- initial respondents selected by probability methods but additional ppl obtained from info provided by initial respondents.