L6 - Collecting and comparing data Flashcards
COMPARING HEALTH OF GROUPS
i) which type of study surveys one group of individuals to test association between exposure and outcome?
ii) which type of study observes populations to est associations between exposure and outcomes?
iii) are the above studies experimental or observational?
iv) which study follows up disease free particpants with certain exposures to see if they develop a disease?
v) which type of study looks at groups who differ at onset on disease status and then look back at exposures?
vi) which type of study randomly allocates groups to receive an intervention?
i) individuals - cross sectional study
ii) populations - ecological study
iiI) observational
iv) cohort study
v) case control study
vi) RCT
CASE CONTROL STUDIES
i) do the groups have different or the same disease condition at the start?
ii) between the two groups - what happens in relation to variables not of interest? and those of interest?
iii) is this a prospective or retrospective study?
iv) what cannot be calculated using this study? why?
v) what can be calculated?
i) different - one will have the disease and one group wont
ii) variables not of interest are matched
- variables of interest are measured and not matched eg exposures
iii) retrospective
iv) cant calculate risk as risk is the probability of developing the disease but you already know if they have the disease
v) can calculate odds ratio - odds of someone having the exposure
CASE CONTROL STUDY CONTINUED
i) what does an odds ratio <1, 0, >1 indicate?
ii) can this study elicit cause and effect relationships? can multiple expsures be identified?
iii) when may this study be good to use? which two things are minimised?
iv) can prevalence/incidence be calculated? how suitable is it for rare exposures?
v) give three disadvantages of this study type
i) OR <1 = exposure may be protective
OR = no assoc between exposure and disease
OR >1 = exposure may be associated with disease
ii) yes can elicit cause and effect
- can identify multiple exposures
iii) good to use when the disease is rare
- minimised selection and information bias
iv) prev/incidence cant be calculated
- less suitable for rare exposures as nobody may have had them
v) disadv - depends on retrospective data availability, hard to find suitable controls, vulnerable to confounding (may be unknown confounders)
RANDOMISED CONTROLLED TRIALS
i) what is it?
ii) name three things they can be used to calculate? and give two examples of each
iii) what does randomisation imply about potential confounders? can exposures related to the disease be causally identified?
iv) give three strengths and three weaknesses
i) a study where participants are allocated randomly between an intervention and control group
ii) calculate safety - safe dose of drug/adverse effects
- efficacy (works in controlled conds) or effectiveness (does it work in the real world?)
- does it work and is it superior to a previous drug, is it cost effective
iii) randomisation means confounders should be equally distributed between the groups
- exposures can be causally identified
iv) strengths = safety and efficacy of new interventions can be established, minimise selection/info bias, best single study evidence for causal associations
weakness = time consuming, not immune to bias, issues wtih participant drop out, patients in trial may not reflect real world groups
TYPES OF DATA
i) what type of variable is a yes no question?
ii) what type of variable is in categories but in no order eg sexual orientation
iii) what variables are in categories that are ordered eg socioeconomic status
iv) which type of variable is on a scale eg number of symptoms?
i) categorical - binary
ii) unordered categorical
iii) ordinal
iv) continous
STATISTICAL TESTS
i) which statistical test should be used for looking at interventions on a continous scale if there are 2 groups? if there are >2 groups?
ii) which test should be used if the variables being compared are categorical?
i) continuous 2 groups = T test
continous > 2groups = ANOVA
ii) categorical = chi squared
ERROR AND POWER
i) what is a type I error?
ii) what is a type II error?
iii) what allows protection against a type I error?
iv) what allows protection against a type II error?
v) what is statistical power? what is it typically accepted at?
i) type I - you observe an outcome but there isnt one
ii) type II - you dont observe a difference but there is one
iii) protect against type I = P value
iv) protect against type II = statistical power
v) power of the study to detect when significant effects are present - typically accepted at 80-90%
P VALUE & CONFIDENCE INTERVALS
i) what is it?
ii) what type of error does it protect against?
iii) how can a P value be decreased?
iv) what do confidence intervals test for? what are they usually given at?
i) probability that the difference observed could have occured by chance
ii) type I error
iii) bigger sample size
iv) test for precision - probability that the true value falls within a range
- usually 95%
T TESTS
i) what are they testing?
ii) when does the between-group difference need to be bigger?
iii) what are the two outputs?
i) whether there is a difference between two variables (can you reject the null hypothesis that there is no diff)
ii) need a bigger between group difference if there is more variability within groups
iii) get a T statistic and a P value
CHI SQUARED TESTS
i) what type of data is this used for?
ii) what are the expected findings?
iii) how does the test arrive at whether there is a difference between groups?
iv) what two outputs does it give
i) categorical eg link between having endometriosis and having ovarian cancer
ii) expected findings are if there was no link between the variables - how many people would be in each group just by chance
iii) compare expected to observed findings and calculate differences
iv) chi squared statistic and p value
CONFIDENCE INTERVALS
i) what are they?
ii) what does it use the current sample to do?
iii) what paramter are used in confidence intervals?
iv) if a confidence interval is 95% - what does this mean for the true values?
v) how do CIs change as sample sizes increase?
iv) what may be inferred if most CIs contain 1? why?
i) range of values that should contain the true population parameter
ii) uses current sample to estimate the actual distribution of a variable in a whole population
iii) odds ratio
iv) CI 95% - CIs will contain the true value 95% of the time and fail to contain it in 5% of the time
v) CIs get smaller with increased sample size
vi) if CIs contain 1 - cant be sure if there is a difference between variables as odds ratio = 1 means no association
SUMMARY
i) what is prevalence? what is incidence?
ii) name three observational studies? what do they do?
iii) what statistical test may be used for continuous data? what may be used for categorical data?
iv) what estimates probability? what estimates precision?
i) prevalence = measures health needs of pop at one time
incidence = measures over time
ii) case control, cohort and cross sectional
- describe populationsa and identify potential causal exposure factors
iii) continuous = T test or ANOVA
catgorical = chi squared
iv) probability = P value
precision = confidence interval
