Stats unit 1

simple random sample

1. list of all members of the population and number off 2 to N

2. use the hat or RNG method

3. Survey the people with the corresponding numbers

hat method

N identical slips of paper, labeled 1 to N. put in the hat and mix them up. draw slips of paper without replacement

RNG method

use RNG to generate unique numbers from range 1 to N

stratified random sample

group individuals by some similar characteristics that’s relevant

take a SRS from each stratum

cluster random sample

group the individuals by location 

number off the clusters and take a SRS of the clusters

survey ALL the people in the randomly selected cluster 

pros: fast, easy cheap

cons: not representative 

Bias

consistent UNDER or OVER estimate of value

Must state direction 

ex. pushups → volunteers were more likely to be fit → OVER estimate 

Bad sampling methods

voluntary and convenience

undercoverage

subset of population can’t be selected bc they weren’t on the sampling frame of all people in the population 

response bias

get an answer but it’s not truthful

nonresponse

1) people don’t respond

2) can’t be reached

Principles of experimental design

1) random assignment

2) control

3) comparison

4) replication - need to have enough experimental units to say it was the treatment that caused the change and not just chance variation from the random assignment 

Randomized block design

Stratified but for experiments

1) experimental units are put into groups (blocks) that share some similar characteristics

2) performs the experiment on all blocks then combine results 

reduces variability, easier to determine if treatment really caused a change 

Who can you generalize the results of your study to? 

Only people SIMILAR to the volunteers in the experiment