1/17
Looks like no tags are added yet.
ways to reduce bias
control group
randomization
blinding
ways to reduce sampling error
replication
balanced design
blocking
matched pair design
extreme treatments
control
treatments should include at least two groups: treatment and control
serves as a baseline to compare the effect of the treatment
randomization
subjects randomly assigned to treatments
includes random sampling and random assignment
random sampling
focuses on who gets into the study
randomly selecting members of a population to participate in a study
random assignment
focuses on who receives which treatment or condition
randomly sorting the sample participants into control and experimental groups
what does random sampling ensure?
ensures sample is representative of the whole population
what does random assignment ensure?
ensures there is no systematic differences between the members of each group
what is blinding?
concealing information from the participants about which experimental unit receives which treatment
single blinded: only subjects are unaware of which subjects receive the treatment
double blinded: subjects and researchers are both unaware of which subjects receive the treatment
what is replication?
each treatment includes multiple independent experimental units (replicates)
what is balanced design?
all treatments have the same sample size
what is blocking?
divides the experimental units into blocks with similar characteristic
steps for blocking:
assign subjects to cohorts based on similarities
randomly assign subjects in each cohort to control and treatment groups, maintaining balance
randomize the order of treatment within each group
what is the primary benefit of blocking?
removes variation between individuals to allow treatment differences to be better estimated
what is matched pair design?
subjects are matched based on key variables, one is assigned to treatment group and one is assigned to the control group
blocking vs. matched pair design
both reduce effect of unrelated variation, but matched pair design is only about two treatments
why is matched pair design a good option for observational studies?
because we cannot do true blocking (randomly assigning participants to treatment or control) in observational studies
matching ensures that comparison groups share critical baseline traits (such as age, health status, socioeconomic background) —> making it more likely that observed differences are caused by the treatment rather than pre-existing differences
what’s extreme treatment?
the use of extreme treatments to help detect treatment effect