Study design and summarising data Flashcards Preview

Epidemiology and biostatistics > Study design and summarising data > Flashcards

Flashcards in Study design and summarising data Deck (23)
Loading flashcards...

what is a RCT

intervention study where subjects are randomly allocated to treatment options


What are the treatment groups compared with in RCT

placebo or uussual care


What does randomisation in RCT allow for

subjects characteristics don't affect which treatment they receive=unbiased
to balance treatment groups by subject characteristics to make sure differences arissing can be attributed to being caused by treatment alone
fair test of efficacy, allows for blinding


what does blinding allow

concealing the allocation. Can be double or single blind


What is ITT and what does it achieve

attributing subject to the group they were originally allocated regardless to whether they change treatment plan.
You can be certain about balance of treatment groups with respect to characteristics of subjects so provides unbaised comparison of subjects


What is per protocol analysis and what negatives does this have, when is it useful

patient analysed according to treatment actually received
negative: balance in patient characteristics present after random allocation is lost.
useful: can be used with ITT if patient stopped/changed treatment


what is the limitation of case control study design

choice of control group affects comparison between case and control
exposure to risk factor usually collected reterospectively and may be incomplete, inaccurate or biased


How does cohort studies differ with case control study

longitudinal and starts with unselected group of individuals who are followed up for set period of time. Used to confim finding of case control


Cohort study design

unselecteed group of healthy individuals->followed up to monitor disease and potential risk factors
prospective, risk factor of disease recorded before disease is confirmed
can be retrospective but requires that full risk factor data is obtained on all individuals with and without disease of interest using data which was recorded prospectively


Difficulty with cohort studies

large no. subjects needed to obtain enough individuals who get disease or conditon particularly if uncommon
length of follow up may be substantial to get enough diseased individuals so cohort study is not feasible for rare diseases
difficulty in maintaining contact with subjects if lengthy follow up
resources required may be high


Cross sectional study what

sample chosen and data on each individual collected at one point in time. Could be different time points for each subject


when is cross sectional study used

survey of prevalence, attitudes/views, inter relationships between variables


why summarise data

data quality monitoring
data checking and data cleaning
baseline data in study
before doing complex analysis


what is ordinal data

data arranged in numerical order (can be quantitative or categorical data)


consequences of categorising continuous data

dichotomising (recategorizing in 2 groups) could be problematic because info and statistical power lost in analysis. Masks nature of relationships
grouping into several groups, stat power effect less than dichotomising. No problem is summary statistic and original data used in analysis
grouping can be useful in nonlinear relationships


what is the arithemetic mean

sum of all values/total no. values


what does standard deviation show

how dispersed the data is


how is variance calculated

summing the squared differences between overall mean and each value and dividing by no.values-1


why is standard deviation better than variance

easier to calculate, same unit as original data


What types of measures for centre of data do you use for: contin data with symmetrical distribution, continuous with positively skewed distribution, continuous data with skewed distribution, discrete data

symmetrical distribution-arithmetic mean
continuous with positively skewed distribution-geometric/harmonic mean
continuous data with skewed distribution-median
discrete data-median unless range of data is large and calculation of mean is sensible


what method to use for spread of distribution

continuous-standard deviation
continuous with skew-IQR


What to include in summary of nominal and ordinal data

nominal-freq and%
ordinal-freq, %, cumulative freq, cumulative %


How to spot skew

positive-tail on right is longer
negative-tail on left is longer