AP Stats U4

4.1- Voluntary response

taken from volunteers

Very biased twoards strong opinions

4.1- Convenience response

taken from easiest to reach, prone to bias

4.1- Simple random sample

Everyone from all groups has an equal chance of being selected.

• label all with #

• choose SRS with random # generator no repeats

• select individuals and survey

4.2- Stratified random sample

Split into stratas of homogeneous characteristics (Some from all strata’s)

• Split into groups

• Run a SRS within each group

• each strata has individuals with shared attributes

4.2- Good estimate

Sampling method that produces low bias and low variability

4.3- Cluster Sample

• The population is divided into clusters (natural/heterogeneous)

• randomly select some clusters

• sample all within those clusters

4.3- Systematic sample

• select a random start point from entire sample

• select a strategic interval

• select every kth individual to sample

4.4-Sampling bias

Bias that occurs when the sampling method tends to favor certain outcomes

4.4-Under coverage

Some individuals are less likely or wont be chosen to be surveyed (Using landlines instead of cell phones to contact people)

4.4- Non response bias

Individuals can’t be reached or refuse to answer (don’t answer, hang up, spam)

4.4- Response bias

Problem with data gathering instrument or process. (People lie, self reporting, whose asking, wording of question, social pressure)

4.5-Oberservational study

A study that is observed with no imposed treatment, often voluntary, or viewing what already exists.

4.5- Experimental study

Researchers deliberately impose a treatment to measure a response.

4.5- Difference between an experimental and observational study

Only experiments can show a cause and effect

4.5-Experimental unit

What is the treatment and who is it imposed on

4.5- Treatment

What is done (Or not done) to the experimental units

4.5- Confounding variables

A variable related to both the explanatory and response variable that makes it hard to determine the cause. 

4.6- Well designed experiment must have…

• Comparison (2 or more treatments)

• Random assignment (Not convenience or voluntary)

• Replication (Multiple in each treatment)

• Control (Control potential → confounding variables)

4.6- Random assignment

Using chance to assign subjects to treatment groups

• Label numbers

• randomize

• assign to show causation

4.6- Placebo affect

• When a fake treatment works

• Blinding (when subjects don’t know the treatment)

• Double blinding 9When experimenters don’t know the treatment)

Difference between random sampling and random assignment

• Random sampling shows generalizations

• Random assignment shows causation to treatment

4.7- Block design

Groups of experimental units that are similar and could affect the results. Then randomly assign treatments within each block.

4.7- Matched pairs design

Subjects are paired based on similarity, and then random assign treatment within each pair, one for each.

Or, both subjects receive both treatments in random order for each subject.

4.8- Simulation

Dot plot with dotted results from a random design using results from experiment.

• Then put actual line on it,

• see how many greater than, how many less than.

• find percentage of how many greater than

• Less than 5% above actual is statistically significant.

4.8- Statistically significant

• When results of an experiment are unlikely (<5%) to happen purely by change

• if statistically significant, we have convincing evidence that the treatment caused the difference in results between treatments

4.9- Cause vs association

Cause - When subjects are assigned to treatments using random assignment, it allows us to conclude that the treatment causes changes to the response variable

Association - Either an observational study, or no imposed treatment

4.9 - Generalization

when there is a random sample of subjects from the population, we can generalize our conclusion to the population from which we sampled from.