Module 0 Flashcards
population parameter
a number that describes something about an entire group or population
ex: census is done to find the population
Census
A sample that includes the entire population
Problems with taking a census
expensive
undercoverage (doesn’t include everyone)
time-consuming
sample statistics
collecting data from a sample to provide a statistic in order to avoid the problems of taking a census
Population inference
results from the sample can be generalized to the entire population
Causal inference
different responses caused by treating two groups differently
random sampling
selecting individuals randomly for a sample.
Can then make population inferences
ensures sample on average looks the same as the population
extraneous factors
variables that are not being tested but could affect the final data to be inaccurate if not controlled
bias
the tendency for a sample to differ from the corresponding population in some systematic way
SRS
simple random samples
A sample chosen randomly, each with equal probability of being selected
ex: drawing names at random from a box
sampling variability
different samples of the population will have slightly different results. These differences are called sampling variability.
stratified random sampling
an SRS is taken from each strata of a population and the results are combined.
strata
homogeneous groups that make up a population
purpose of stratified random sampling
reduce bias and reduce variability
systematic random sampling
start from a randomly selected individual and sample every kth person
should be no reason for the order of the list that could alter the results
less expensive
cluster random sampling
splitting the population into similar groups (clusters), selecting one or a few, and perform a census on each chosen cluster
if each cluster fairly represents the population, gives an unbiased sample
selection bias (undercoverage)
when a portion of the population is not sampled or has a smaller representation than in the population
response bias
the survey design influences responses to give a biased result
voluntary response bias
people can choose if they want to participate in the sample
nonresponse bias
when a large proportion of the sample fail to respond
random allocation
randomly selecting individuals for each group to reduce lurking variables
lurking variables
variables related to group membership and the response. in causal inference, groups should be selected using random allocation so lurking variables do not effect the response.
observational study
the investigator observes individuals but does not attempt to influence responses
retrospective study
choses sample based on if individuals meet the criteria for what study they are conducting. Ex: only choses participants near powerplant
prospective study
choses subjects for the sample first and then observes outcomes
randomized, comparative experiments
can prove cause and effect relationship
randomly selects sample
randomly assigns sample to treatment groups
treats 2 groups differently