Statistics Flashcards
what is the null hypothesis?
this is the statement that there is no relationship between two measured phenomena
how do we determine the NNT?
1/ARR
on a 2 x 2 table, what represents the Type I error and what is Type II error?
Type I error is when the test suggests there is a difference, but in fact there is NO difference.
Another way to frame this would be that the test incorrectly rejects the null hypothesis
this is what p value reflects
type II is when a test incorrectly suggests there is no difference, but in fact, there is.
This is a false negative. The test does NOT reject the null and is INCORRECT
what is statistical power?
1 - beta-error
this could also be worded as:
1 - chance of incorrectly failing to reject null
what sort of study measures incidence?
what sort of study measures prevalence?
incidence is a cohort study
prevalence is just a prevalence study
what is “attributable risk”?
how is it calculated
this is the incidence of disease attributable to exposure
AR = I(exposed) - I(nonexposed)
what is teh calculation for population attributable risk?
what is the calculation for population attributable fraction?
pop attrib risk is the incidence of a disease in a population associated with the occurrence of a risk factor
AR(p) = AR x P(risk factor)
Pop attrib fraction = what fraction of disease in a population is attributable to exposure to a risk factor
AF (p) = AR(p)/I(total)
what is the difference between an efficacy trial and an effectiveness trial?
the efficacy is a treatment under ideal conditions
effectiveness = treatment under ordinary conditions
what is the role of intention to treat?
this is a way to minimise bias
basically you have to analyse people based on whether you’d planned on treating them. NOT based on whether they took the drug/intervention
what is a case control study?
what sort of information does a case control study yield?
this is a retrospective study looking at an issue possibly leading to a a particular condition
it provides an odds ratio. This is done because it is not possible to measure incidence.
how do we calculate an odds ratio?
odds ratio = odds that a case is exposure / odds that a control is exposed
the formula is: AD/BC
in words, what does a positive likelihood ratio measure?
it is the proportion of patients with disease with a positive test, divided by the proportion of patients WITHOUT disease with a positive test.
that is,
[a/(a+c)] / [b/(b+d)]
also:
sens/ (1-spec)
(a neg LR is (1-sens)/spec)
if you are looking at a study and there’s no “time variable” what is the way to determine whether someone will have the disease?
WOW! That was a convoluted way of asking about Odds ratios.
The way that we run meta-analyses is by odds ratios. This is so we can compare studies over different time periods
that’s usually what those “forrest plots” show in Cochrane reviews
what is the meaning of the different levels of evidence?
what is level I
II?
IV?
I - at least one systematic review multiple good RCT’s
II - at least one good sized RCT
III - well designed trials without
randomisation - pre/post, cohort, time series, matched case control
IV - - non-experimental studies from more than one research group
V - Respected opinion
what is the best kind of trial to figure out the dose response of a treatment
what about reversibility of a condition
what about causation?
Temporality - Clinical Trial
Strength - cohort study
Dose-Response - case-control
Reversibility - cross sectional
Consistency - aggregate risk
Biological Plausibility - case series
Specificity - case report
