Ap Statistics - Planning a Study

unit 2

Voluntary response sample

Overrepresents strong opinions because individuals self-select themselves into the sample

Simple random sample

Every possible group of the given size has to be equally likely to be selected —> all combinations are possible (ex: lottery)

Stratified sampling

The population is divided into subgroups (called strata), and a random sample from each group creates the sample.

Sampling error 

• Natural variation between samples (not an error committed by any person)

• Can never be eliminated

• Can be described using probability

• Generally smaller if the sample size is larger.

Poor survey design or faulty sampling techniques

A large sample size cannot account for…

Use stratified random sampling

If there is still high variation…

Random sampling error

occurs due to chance variation/luck

Sampling method error

occurs due to faulty sampling method choice (ex: quota, non-probability sampling)

Non-sampling method error

occurs in the members’ responses (ex, human error (kg vs. g), dishonesty) 

Characteristics of a Well-Designed, Well-Conducted Survey

• Random assignment

• Replication

• Control

Selection bias

a sample selection that does not accurately reflect the target population —> biased conclusions

Nonresponse bias

When not all individuals are contactable, even if they are randomly selected

Unintentional bias

When the surveyor tries to systematically pick people representative of the whole population.

Convenience samples

Units are conveniently selected into the sample (ex: asking friends)

Response bias

Participants respond inaccurately to questions due to:

• Wording of questions

• Honesty

• Order of choices

• Demeanor of the interviewer

Household bias

Occurs when one type of respondent is overrepresented because groupings of different sizes are polled equally

Observational studies

Proves correlation through surveying & observing data collection

Experimental studies

Proves causation from control VS treatment groups

Parameter

Characteristics of a population

Census

Contacts every individual in the population (ex: government surveys)

Undercoverage bias

When part of the population is excluded from the sample (ex: door-to-door surveys ignore the homeless)

Systemic sampling

Every k^th unit is sampled from an ordered list of the population

Cluster sampling

Population divides into subgroups (called strata) under a characteristic, then a random sample from each strata composes the sample.

Multistage sampling

Using a combination of different sampling methods

Confounding

A type of lurking variable.

• When the investigator is unable to separate their respective effects on the response variable.

Lurking variable

An (explanatory) variable that influences the response variable but isn’t measured or studied in an experiment.