lecture 22 Flashcards
what is internal validity
whether the study design, conduct, and analysis answer the research questions without bias
what is external validity
whether the findings can be applicable to broader populations, also known as generalisability
what does increasing the size of a sample when random sampling do
- reduces sample variability
- increases likelihood of getting a representative sample
- increases precision of parameter estimate
what is estimating population parameters
when you use the findings from your study population to estimate what would occur in the source population
what are the 2 methods of measuring the influence of sampling error
confidence intervals
p values
how does a 95% confidence interval work
if your repeated a study 100 times with a random sample each time, 100 estimates and 100 confidence intervals
then….
- 95 of them = the parameter would lie within that studies 95% confidence interval
- 5 of them = the parameter would not lie within the studies 95% confidence interval
what is an interpretation of what a confidence interval is
we are 95% confident that the true population value lies between the limits of the confidence interval
what does a smaller confidence interval mean
more precise
what does increasing the sample size of a study do to the confidence interval
increasing the sample size can make the confidence interval narrower
what can a confidence interval help us decide about a study
whether it is clinically important
what is the estimate or point estimate
the measure found in the study sample
what is the parameter
the true value of the measure in the population that the study is trying to discover
what do confidence intervals give us a sense of
how precisely we are estimating the population parameter
what are p vales
probability of getting study estimate (or one further from the null) when there is really no association, just because of sampling error (chance)
what logic do p values use
logic of hypothesis testing
in the 2 by 2 table of association vs no association what do p-values tell us about
when we find association in the study results and there is no association in the parameter
what is the null hypothesis
when there really is no association in the population
when the parameter equals null values the ratio and difference measures equal ….
odds ratio and relative risk = 1
risk difference = 0
what is the alternative hypothesis
when there really is an association in the population
when the parameter does not equal the null value the ratio and difference measures will equal ….
odds ratio and relative risk = will not equal 1
risk difference = will not equal 0
what is the symbol for null hypothesis
Ho
what is the symbol for alternative hypothesis
HA
example of interpreting a p value when the …
odds ratio = 2.3
p = 0.03
the probability of finding an OR of 2.3 (or further from the null) when the null hypothesis is true is 0.03 (or 3%)
what is the threshold for p values
0.05
what is said if the p value is less than 0.05
- reject Ho
- accept HA
association is ‘statistically significant’
what is said if the p value is greater than 0.05
- fail to reject Ho
- reject HA
association is ‘not statistically significant’
interpretation of p value when it is less than 0.05
since the p value is less than 0.05 the association is statistically significant. chance is an unlikely explanation of the study finding
interpretation of a p value when it is greater than 0.05
since the p value is greater than 0.05 the association if not statistically significant. the study finding is consistent with chance as an explanation
what are type II errors
when incorrectly fail to reject Ho when should have
(p should have been < 0.05 but got > 0.05)
what are type II errors usually from
typically due to having too few people in the study
what does sample size mean for the p value
bigger sample size = more likely to get small p value
smaller sample size = less likely to get small p value
what is said if the 95% confidence interval includes the null value
p > 0.05
not statistically significant
what is said if the 95% confidence interval doesn’t include the null value
p < 0.05
statistically significant
how should you report p values
useful to report p values not just statistically or not statistically significant
why are p values problematic
- arbitrary threshold
- only about Ho = just gives evidence about consistency with the null hypothesis
- nothing about importance
what should be considered when evaluating internal validity
- chance = confidence intervals
- bias = strengths and limitations
- confounding = comparing analysis phase (like multivariable analysis etc)