chapter 4: designing studies

simple random sample (SRS)

every group of size n has an equal chance of being chosen

stratified random sample (homogenous grouping)

splits population into groups of similarity (strata), then take an SRS of each group; those chosen from each SRS will be selected for the sample

cluster sample (heterogenous grouping)

splits population into groups based on location (clusters), randomly select group, survey everyone in that group

difference between strata and clusters 

cluster-mini population all located together, strata-group should influence the choice 

systematic sample

randomly select a starting point and every nth individual thereafter will be selected for the sample

undercoverage bias

when members of the population are less likely to be chosen for a sample

non-response

individuals selected to be in a sample but can’t contracted or refuse

response bias

inaccurate responses (lying or confusing questions)

population

all items or subjects

sample

subset of population

census

selection of all items/subjects in a population

question wording bias

questions that are confusing or leading

observational study

no treatments imposed/no random assignment, only observed (NO CAUSE/EFFECT)

retrospective

examine existing data

prospective

track individuals into the future

experiment

treatments imposed

experimental units

individuals assigned treatments (humans often called subjects)

explanatory variable

variable whose levels are manipulated intentionally

response variable

outcome from the treatments administered

confounding variables

variables not controlled that affect results

well designed experiment (2 C’s, 2 R’s)

control group, control confounding variables, random assignment, replication

control group

compares two or more treatment groups

control confounding variables

variables that may affect results

random assignment

allows causation and balances out the effects of confounding variables

replication

use enough experimental units to distinguish differences

random selection

generalizes population

random assignment

allows causation

control group

a group receiving an inactive treatment (placebo)

placebo effect

when experiment units have a response to a placebo (fake treatment works)

randomized block design

at the beginning of the experiment, units are divided; block: group of experimental units that are similar in some way

matched pairs design

a pair of similar experimental units that are matched then randomly assigned to each treatment

statistically significant

when results from a study are too unusual to have occurred purely by chance, <5%