Korea Social Survey Analyst Grade 2 common mistakes
Korea Social Survey Analyst Grade 2 · Understand the traps and how to avoid them.
Calling a large opt-in poll representative
Self-selection and missing population groups can remain even with many responses.
What to do instead
Evaluate the selection mechanism and coverage before making a population claim.
Combining two policy questions in one item
A respondent can agree with one component and disagree with the other.
What to do instead
Separate the concepts and test each question’s wording.
Inferring a treatment effect from correlation alone
Uncontrolled causes, reverse direction and selection can produce the observed association.
What to do instead
State the study design and distinguish an association from a justified causal claim.
Coding every missing response as zero
Zero can be a real value, while refusal and nonapplicability have different meanings.
What to do instead
Use documented separate codes and an explicit analysis rule.
Editing an unusual answer without evidence
A surprising observation may be genuine and changing it can introduce bias.
What to do instead
Check the source or respondent when appropriate and retain an audit trail.
Assuming weighting fixes every missing group
The adjustment depends on observed information and cannot recover all unmeasured differences.
What to do instead
State the weighting variables and remaining nonresponse limitations.
Calling the p-value the probability a hypothesis is true
The probability is calculated conditional on the specified null model, not over the hypothesis itself.
What to do instead
State the null model and describe extremeness of data under that model.
Treating a small p-value as a large practical effect
Sample size and variability influence a test and do not define practical importance.
What to do instead
Report the estimated effect and uncertainty in its substantive context.