chance Flashcards
whats the crux of internal validity
is our study estimate an accurate estimation of the actual value in the source population
whats the crux of external validity
the extent to which the study findings are applicable to a broader or different population
aka generisability
define chance or sampling error
when we repeatedly sample randomly from a source population and most of the time we get a representative sample, chance is the random times where we do not get a representative sample randomly.
what can be done to reduce chance, but whats the catch of this
we will take larger sample sizes inorder to reduce sampling error. but we cannot fully eliminate sampling error
increasing the sample size has what 3 effects
reduces variability in the sample, increases likelihood of getting a representative sample
increases precision of parameter estimate
whats the key catchphrases for 95% confidence intervals
the exposed group were X times as likely to develop the outcome than the comparison group
we are CI interval number (95%) confident that the true value of the MoA lies between x and y
this is/isn’t a statistically significant finding at the 0.05 level as the CI does or does not include the null value of X. chance is thus an unlikely/likely explanation of the study finding.
confiedence intervals can also help us determine if MoA is
clinically important
define estimate and parameter
parameter is the true value of the measure of association in the population that the study is trying to discover
estimate is the measure of association found by the study, can be referred to as the point estimate
to be clinically important the finding has to be what
the confidence interval has to be below the predetermined mark of clinical importance
what are p values
the probability of getting a study estimate or one further from the null, when there is not one, just by chance
p value tells us about what scenario
when we found an association but there is not really one present
whats the null hypothesis
there really is no association, MoA is null value
whats the alternate hypothesis
when there really is an association in the population, MoA does not equal null
whats a type one error and whats the threshold
type one error is what the p value tells us about. its the probability of finding an association when there really isn’t one.
threshold is 0.05
whats the magical 3 lines for interpreting a p value
the exposed group were X times as likely to develop the outcome compared to the comparison group.
the probability of a MoA of X or one further from the null, when the null hypothesis is true, is P value
As the p value is above/below the value of 0.05 this is/is not a statistically significant finding. chance is therefore a likely/unlikely explanation for the finding of the study.
whats a type two error
when we find no association where we should have, when we incorrectly fail to reject Ho when we should have
how do we reduce the type 2 errors
by not having too few people in the study
what is calculated to see how to reduce type 2 error
power will be calculated to see how many people are needed in the study to remove type 2 error.
why are p values problematic
arbitrary threshold, says nothing about importance and only talks about Ho
explain arbitrary threshold
p value was decided on arbitrarily, like is p of 0.04 different to 0.06. thus we report p values but don’t want to use them to say if significant or not
explain says nothing about importance
only gives evidence of consistency with the null hypothesis, doesn’t say anything about precision
explain nothing about importance
stastical significance is not clinical importance, doesnt tell us if findings are useful or not.
lack of statistical significance is not evidence for no association.
