Evaluate statistically based reports NCEA Level 3

Many questions, may have branches and skips, may have tick multiple boxes option, may have open ended write a comment questions

Surveys

Few questions, multi choice only

Polls

Each sample taken from a population will have a different result.

Sampling error

The difference between the results obtained from a sample survey and the result that would've been obtained from the whole population

Sampling error

Errors that occur in the collection, recording and tabulation of data due to poor sample design, mistakes in collecting and recording data and behaviour of people surveyed. Are always present.

Non-sampling error

A sample drawn in a way that each element of the population has an equal chance of being selected. Likely to be representative of the population.

Random sample

A sample selected in a way that each member has the same chance of being included in the sample. Time consuming - best suited to small populations.

Simple random sample

Starting at a random point and sampling every nth individual. More convenient than simple random. Likely to be representative of population.

Systematic random sample

Divide population into groups called strata. Then a sample is selected from each of the strata in proportion to size of group. Collection of all samples from all strata.

Stratified random sample

Stratified sample that may or may not be representative of the population. Researcher keeps sampling until they have enough of each category being investigated.

Quota sample

Whole population divided into geographical groups. Each cluster is representative of the population. Random sample is taken from each group.

Cluster sample

Sample taken from easiest part of population to reach. May or may not be representative of population.

Convenience sampling

Open to anyone + people can decide if they partake in the survey. Likely to be biased.

Self-selected sampling

Cost is low, response rate is poor - biased towards educated people + people with interest in topic, no knowledge about non-response, long time between data collection and analysis, geographic distribution can be wide.

Mailed questionnaire

Most used, numbers selected randomly from a phone book or by generating random numbers, unlisted numbers, landline coverage is decreasing - increasing bias towards young people, geographic distribution can be wide, data collection period short, short + simple questions.

Telephone survey

High cost, data collection long, geographic distribution must be clustered, good rapport, long time, control and quality over responses, sensitivity issues.

Face-to-face surveys

Low cost, quick data collection, geographic distribution wide, questionnaires may be complex, more fun and dynamic (videos etc), if self selected results may be biased.

Online surveys

Occurs when sampling frame, list sample is taken from, does not include many members of the population. Sampling frame is not representative of population.

Selection error

Occurs when many people in a sample do not respond. Distorts results.

Non-response error

Potential bias if non-respondents are likely to behave differently to respondents with respect to the question being asked.

Non-response bias

Occurs when people do not give the correct response to the question in a survey. Anything in the survey design that influences the result - wording of questions, interviewer influencing response, sensitive question etc.

Response error

Caused if many members of the population do not respond to a survey because it is voluntary. Usually only people with strong opinions will respond.

Voluntary response error

People decide themselves whether to be surveyed or not.

Self-selection bias

Subtle variations in wording influences responses.

Question effects

People tend to answer questions in a way they consider to be socially acceptable.

Behavioural considerations

Different interviewers asking same question can obtain different results - influence how people respond.

Interviewer effects

Question order, survey layout

Survey format effects

Taking the data from one population and transforming the results to another.

Transferring findings

How people select people to survey

Sampling methods

How people get the information from the people

Survey methods

A range of values that we're pretty sure the population parameter lies within.

Confidence interval

Gives us an indication of the maximum likely size of the sampling variability and allows us to put a bound around where we think the population parameter will be.

Margin of error

Confounding variables, extending results inappropriately, using the past as a source of data - recalled information often unreliable

Problems with observational studies

Causal claim cannot be made

Observational studies

Causal claim can be made if random allocation occurs

Experiments

Random allocation, use of control group, use of placebo, use of blinding

How to design a good experiment