1/18
Looks like no tags are added yet.
Name | Mastery | Learn | Test | Matching | Spaced |
---|
No study sessions yet.
What is the difference in observational studies vs randomized trials in terms of cost
observational studies are generally less expensive but sometimes run for a long time,
randomized trials require more resources but can determine causality and control for confounding variables
Who is in a randomized trial?
the study population must be representative of the population we want to generalize the results to, ideally we randomly select who is in the study to remove selection bias and randomize which treatment they get
What data is collected in a randomized trial?
exposure (treatment groups) assigned and received, outcomes (improvement, side effects, mortality), demographic information/prognosis profile (ie. are older individuals more likely to have adverse outcomes), information about masking/blinding (did the participants know which treatment they were receiving, did participants suspect they were getting a placebo
What is standardized randomization?
When certain characteristics could potentially be confounders we represent those in the treatment and control groups (aka blocking) so each subgroup has its own mini-experiment and we are able to have the effect of treatment on each group assessed separately
What is an intentional crossover?
When participants intentionally switch treatment groups at some point in the experiment, and individuals serve as their own controls (however this is not possible if either treatment is curative)
What is an unintentional crossover?
when participants are assigned to one treatment but receive the other treatment instead, causes significant issues in analysis
What is bias from noncompliance?
a type of selection bias that occurs when individuals who choose to not comply with the study protocol differ from those who comply (ie. dropouts, dropins)
What is a type I Error
aka false positive, example would be when an innocent person is convicted or we conclude a treatment is effective when it really isn't or we conclude a difference between two treatments when there is no difference
What is a type II Error
aka false negative, example would be when a guilty person is not convicted, when we conclude the treatment has no effect but is actually effective, we conclude that there is no difference between two treatments but one is more effective than the other
What is the probability of type I error
aka alpha, is the probability that we will detect a difference/effect when there is no difference/effect
What is power of a study?
the probability of detecting a real difference/effect = the probability of correctly rejecting the null hypothesis
What is the formula for probability of type II error
beta = 1 - power
What is the multiple testing problem
aka data dredging or p-hacking, the more tests that we do, the higher chance that at least one Type I error will occur, and you will not know which test(s) resulted in type I error
What is the effect of sample size on Type I & Type II error?
larger sample sizes decrease both type 1 and 2 error and ensure that the results have statistical significance, however larger sample sizes require more resources and higher costs
What is internal validity?
the degree to which confounding variables can be controlled, and how well the study is conducted. How well does the study rule out alternative explanations/confounders in the findings
What is external validity?
the degree to which results can be generalized to the world at large; threats to external validity are present automatically if there are threats to internal validity
What is the effect of confounders on internal validity?
they can threaten internal validity, thus why we need to randomize treatments
What are some sources of bias that can happen during the study?
non-response bias, misclassification bias, drop outs
What are some sources of bias that happen after the study?
it's more likely that results that show a treatment is effective are published than results that show that a treatment has no benefit/effect