Unit 3 - Collecting Data

A sample of people that are easy to reach WITHOUT any randomizing
-Asking students to volunteer to answer the survey
-Social media surveys
-Street surveys

Choosing a group from the population so that every individual + every possble group of individuals has a chance of being chosen
-Picking names from a hat

1) label individuals in population
2) randomize
3) select sample

Groups the population by some variable (homogenous groups) and then chooses a SRS from each group
-Group doctors at a hospital by the floor they work on

One group

Multiple groups

Splits population on group based on location (heterogenous group), then randomly selecting EVERYONE from ONE of the groups

RANDOMLY choose which place to start, then sample every nth individual after that

Collecting data from entire population

People CHOOSE to be in a sample (convenience sample), some in the population may not choose to be surveyed --> data isn't reflective of population
-Sample may be different from the people who chose not to respond
-Ex. online survey about running, only runners who like it would respond

When some members of a population CAN'T be chosen/are unlikely to be chosen
-Ex. only surveying people from your grade on what their GPAs are

When an individual is selected to be in the sample but chooses not to participate/cannot be reached AFTER being chose
-Ex. person chosen in a phone survey cannot be contacted after five calls

When there is a pattern of inaccurate responses due to:
-Question wording
-The interviewer
-Individuals lying

No treatment imposed, we only observe

What is done to participants (ex. drinks a different kind of coffee)
-Only assignment of treatment needs to be randomized in an EXPERIMENT

Treatments are imposed and they allow for CAUSATION

What or who the treatment is imposed upon in an experiment

Used to predict outcomes, unit the treatment is imposed upon
-If there is more than one __________ we all them "factors" (ex. type of sugar)
-Factors can have "levels" (ex. 1 tsp of sugar vs 1 tbsp of sugar)

The outcomes of the experiment/observational study

Variables that are not controlled that may affect the outcomes

Treatments

Used to provide baseline data for comparison
-Often given a placebo

Test subjects are unaware of waht treatments were assigned

When neither the test subjects nor the experimenters are aware of what treatment was assigned (ex. because doctors get paid for if their medical trial is a success)

When a fake treatment works

1. Comparison (at least two different treatments)
2. Random Assignment (need a chance process of assigning treatments)
3. Control (keep all other variables besides treatments constant, not always a control group)
4. Replication (use enough experimental units to distinguish differences)

-Separate the experimental units into groups by some similar characteristic (GPA, gender) into blocks
-Then randomly assign treatments within each block

Each group of experimental units that are known to be similar

An experiment used to compare two treatments that use blocks of two test subjects/experimental units
-Two very similar experimental units are paired and then randomly assigned to treatment
-Form of randomized block design
-Ex. flipping a coin to decide if a person tests a new chair or an old chair first
-One participant is randomly assigned to a treatment and the other half of the pair is automatically sent to the other treatment

-Different samples yield different estimates
-Larger samples provide more accurate estimates

Creates an interval of plausible values
-Sample estimate does not equal margin of error

-When the results from the study are too unusual to have occurred purely by chance
-If the probability is less than or = to 5% = statistically significant

Decisions from the sample can be attirbuted to the distribution (population) from which the sample was drawn