STAT Unit 4

population

entire group you want to study

sample

Smaller group taken from the population to study

Convenience sample

type of bias - taking data that’s easy to reach

Voluntary Response

type of bias - people choose to respond - only people with strong opinions tend to respond

Simple Random Sample (SRS)

lowers bias - makes sure there is a equal representation for population

Steps for Simple Random Sample

1. Label everyone

2. Use a random number generator or draw names

3. Select without repeats

Stratified Random Sample

splits population into homogenous groups (share same characteristics. “Sample some from all groups”

Steps for Stratified Random Sample

1. split population into homogenous groups

2. number individuals within each group (1 to N)

3. use a random number generator to select people (no repeats)

4. repeat for each group

Cluster Sample

splits population into heterogenous groups (different characteristics)

“Sample all from some groups”

quick and easy to collect data - good when groups are mixed

Cluster Sample

Systematic Random Sample

1. Label everyone 1-N

2. Choose a random starting point

3. Pick every kth person until sample is complete

reduces variability and ensures groups are represented 

Stratified Random Sample

spreads sample evenly through population

Systematic

Undercoverage

Some groups in the population are less likely to be chosen or not represented.

Nonresponse

people are selected for the sample but do not respond

Response Bias

when survey design or wording influences responses, or people lie

Explanatory Variable

the cause - what you change

Response Variable

effect/ outcome measured

Confounding Variable

hidden variable that affects both explanatory and response variables (messes up results)

Observational Study 

researchers just observe - no treatment given

Experiment

researchers impose treatment - shows causation

Experimental Units

who the experiment is being done on

Control Group

group that does not get treatment - used to show cause and effect

Placebo

fake treatment given to compare effects

Steps to Design a Good Experiment

1. Randomly assign subjects to groups

2. One group gets treatment

3. Other group gets no treatment (control group)

4. Compare results

Random Assignment

randomly putting subjects into groups (RNG)

1. Label

2. Randomize

3. Assign

Blinding

reduces bias

Single Blind 

subjects don’t know - researchers know

Double blind

subjects and researchers don’t know - best way to avoid bias

4 Parts of a Good Experiment CRRC

1. Comparison - use 2 or more groups

2. Random Assignment - randomly assign subjects to groups

3. Replication - enough subjects in each group

4. Control - control group keep other variables same

Completely Randomized Design

all subjects are randomly assigned to treatment groups

no grouping/blocking beforehand

use when subjects are similar

Randomized Block Design 

• subjects are first grouped by a variable that may affect the response variable (block), then randomly assigned to treatments within each block.

• used for confounding variables - reduces variability

• use when groups differ (age, gender)

Matched Pairs Design

• each block has 2 subjects, or each subject gets 2 treatments

• reduces variability

Stimulation

a way to model what could happen by random chance

• repeated random trials to model chance

Statistical Signifigance

• a result is statistically significant if it is unlikely to happen by chance alone

• less than 5%

Random Sample

People are randomly chosen from a population → lets us generalize

Random Assignment

People randomly put into groups → allows cause & effect