Concerns with Sampling and Observational Studies (Exam 1)

Question 1

selection bias

Answer 1

• occurs when individuals in the target population are not sampled at an equal rate
• even when a random sampling design is used, there are other potential sources of bias in the sampling method
• biases can happen at other steps… not just when we’re selecting our sampling

Question 2

perfect coverage

Answer 2

• when the sampling frame lists everyone in the target population (no missing or extra people)
• only sometimes achievable

Question 3

undercoverage

Answer 3

occurs when a certain group of individuals in the target population is not included in the sampling frame and therefore given no chance of selection in the sample

Question 4

overcoverage

Answer 4

occurs when a certain group of individuals not in the target population is included in the sampling frame

Question 5

nonresponse

Answer 5

• occurs when individuals selected for the sample can’t be contacted/refuse to participate in the study
• important to consider WHY these individuals don’t respond… classify nonresponse into 3 explanations
• depending on the nonresponse mechanism we assume to be true, there may (not) be selection bias introduced to the sample, and we may (not) be able to correct this bias

Question 6

missing completely at random (MCAR)

Answer 6

• when the probability of nonresponse is the same for all individuals in the sample
• does NOT introduce selection bias (missing respondents are random)
• since we assumes that the nonrespondents are not different in any way from the respondents, the sample of respondents should still be representative of the target population, so we can remove the nonrespondents from our sample

Question 7

missing at random (MAR)

Answer 7

• when the probability of nonresponse is associated with KNOWN and measured factors
• there IS something special about you about why you’re missing
• assumes that certain groups in the population are more likely to produce nonresponse than others (if these groups are known prior to sampling, then we can oversample them to guarantee proper representation or after sampling, we can adjust the weights assigned to individuals in the sample to guarantee proper representation

Question 8

missing not at random (MNAR)

Answer 8

• when the probability of nonresponse varies for reasons that are UNKNOWN to us, particularly the level of the variable of interest
• no way to verify why these people are missing
• since we don’t know the cause of nonresponse, very difficult to make adjustments/corrections for this source of missing data

Question 9

weighting

Answer 9

• procedure that allows us to place a higher/lower emphasis on certain individuals in the sample
• we want the weighted individuals to add up proportionally

Question 10

what are 3 ways in which selection bias can be introduced into the data collection process?

Answer 10

1) from the construction of the sampling frame (undercoverage & overcoverage)
2) from the methods of selecting individuals from the sampling frame (convenience sampling & voluntary response sampling)
3) from nonresponse

Question 11

experimental study

Answer 11

deliberately imposes some treatment on individuals in order to observe their responses

Question 12

observational study

Answer 12

• observes individuals and measures variables of interest, without controlling any factors that might influence the response
• ex: surveys/polls/questionnaires, interviews, examinations

Question 13

response bias

Answer 13

• occurs when people answer falsely/inaccurately, sometimes due to unintentional behavior by the interviewer
• ex: respondents may lie, especially about illegal/unpopular behavior